Initial commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+ckpts/ filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,419 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+import os, sys
+sys.path.append(os.getcwd())
+
+# Gradio demo, comparison SAM vs ZIM
+import os
+import torch
+import gradio as gr
+from gradio_image_prompter import ImagePrompter
+import numpy as np
+import cv2
+from zim import zim_model_registry, ZimPredictor, ZimAutomaticMaskGenerator
+from segment_anything import sam_model_registry, SamPredictor, SamAutomaticMaskGenerator
+from zim.utils import show_mat_anns
+
+def get_shortest_axis(image):
+    h, w, _ = image.shape
+    return h if h < w else w
+
+def reset_image(image, prompts):
+    if image is None:
+        image = np.zeros((1024, 1024, 3), dtype=np.uint8)
+    else:
+        image = image['image']
+    zim_predictor.set_image(image)
+    sam_predictor.set_image(image)
+    prompts = dict()
+    black = np.zeros(image.shape[:2], dtype=np.uint8)
+
+    return (image, image, image, image, black, black, black, black, prompts)
+
+def reset_example_image(image, prompts):
+    if image is None:
+        image = np.zeros((1024, 1024, 3), dtype=np.uint8)
+
+    zim_predictor.set_image(image)
+    sam_predictor.set_image(image)
+    prompts = dict()
+    black = np.zeros(image.shape[:2], dtype=np.uint8)
+
+    image_dict = {}
+    image_dict['image'] = image
+    image_dict['prompts'] = prompts
+
+    return (image, image_dict, image, image, image, black, black, black, black, prompts)
+
+def run_amg(image):
+    gr.Info('Checkout ZIM Auto Mask tab.', duration=3)
+    zim_masks = zim_mask_generator.generate(image)
+    zim_masks_vis = show_mat_anns(image, zim_masks)
+    
+    sam_masks = sam_mask_generator.generate(image)
+    sam_masks_vis = show_mat_anns(image, sam_masks)
+    
+    return zim_masks_vis, sam_masks_vis
+    
+    
+def run_model(image, prompts):
+    if not prompts:
+        raise gr.Error(f'Please input any point or BBox')
+    gr.Info('Checkout ZIM Mask tab.', duration=3)
+    point_coords = None
+    point_labels = None
+    boxes = None
+
+    if "point" in prompts:
+        point_coords, point_labels = [], []
+
+        for type, pts in prompts["point"]:
+            point_coords.append(pts)
+            point_labels.append(type)
+        point_coords = np.array(point_coords)
+        point_labels = np.array(point_labels)
+
+    if "bbox" in prompts:
+        boxes = prompts['bbox']
+        boxes = np.array(boxes)
+
+    if "scribble" in prompts:
+        point_coords, point_labels = [], []
+
+        for pts in prompts["scribble"]:
+            point_coords.append(np.flip(pts))
+            point_labels.append(1)
+        if len(point_coords) == 0:
+            raise gr.Error("Please input any scribbles.")
+        point_coords = np.array(point_coords)
+        point_labels = np.array(point_labels)
+
+    # run ZIM
+    zim_mask, _, _ = zim_predictor.predict(
+        point_coords=point_coords,
+        point_labels=point_labels,
+        box=boxes,
+        multimask_output=False,
+    )
+    zim_mask = np.squeeze(zim_mask, axis=0)
+    zim_mask = np.uint8(zim_mask * 255)
+
+    # run SAM
+    sam_mask, _, _ = sam_predictor.predict(
+        point_coords=point_coords,
+        point_labels=point_labels,
+        box=boxes,
+        multimask_output=False,
+        )
+    sam_mask = np.squeeze(sam_mask, axis=0)
+    sam_mask = np.uint8(sam_mask * 255)
+    
+    return zim_mask, sam_mask
+
+def reset_scribble(image, scribble, prompts):
+    # scribble = dict()
+    for k in prompts.keys():
+        prompts[k] = []
+
+    for k, v in scribble.items():
+        scribble[k] = None
+
+    black = np.zeros(image.shape[:3], dtype=np.uint8)
+
+    return scribble, black, black
+
+def update_scribble(image, scribble, prompts):
+    if "point" in prompts:
+        del prompts["point"]
+
+    if "bbox" in prompts:
+        del prompts["bbox"]
+    
+    prompts = dict() # reset prompt
+    scribble_mask = scribble["layers"][0][..., -1] > 0
+
+    scribble_coords = np.argwhere(scribble_mask)
+    n_points = min(len(scribble_coords), 24)
+    indices = np.linspace(0, len(scribble_coords)-1, n_points, dtype=int)
+    scribble_sampled = scribble_coords[indices]
+
+    prompts["scribble"] = scribble_sampled
+    
+    zim_mask, sam_mask = run_model(image, prompts)
+
+    return zim_mask, sam_mask, prompts
+
+
+def draw_point(img, pt, size, color):
+    # draw circle with white boundary region
+    cv2.circle(img, (int(pt[0]), int(pt[1])), int(size * 1.3), (255, 255, 255), -1)
+    cv2.circle(img, (int(pt[0]), int(pt[1])), int(size * 0.9), color, -1)
+
+
+def draw_images(image, mask, prompts):
+    if len(prompts) == 0 or mask.shape[1] == 1:
+        return image, image, image
+
+    minor = get_shortest_axis(image)
+    size = int(minor / 80)
+
+    image = np.float32(image)
+
+    def blending(image, mask):
+        mask = np.float32(mask) / 255
+        blended_image = np.zeros_like(image, dtype=np.float32)
+        blended_image[:, :, :] = [108, 0, 192]
+        blended_image = (image * 0.5) + (blended_image * 0.5)
+    
+        img_with_mask = mask[:, :, None] * blended_image + (1 - mask[:, :, None]) * image
+        img_with_mask = np.uint8(img_with_mask)
+
+        return img_with_mask
+
+    img_with_mask = blending(image, mask)
+    img_with_point = img_with_mask.copy()
+
+    if "point" in prompts:
+        for type, pts in prompts["point"]:
+            if type == "Positive":
+                color = (0, 0, 255)
+                draw_point(img_with_point, pts, size, color)
+            elif type == "Negative":
+                color = (255, 0, 0)
+                draw_point(img_with_point, pts, size, color)
+
+    size = int(minor / 200)
+
+    return (
+        img,
+        img_with_mask,
+    )
+
+def get_point_or_box_prompts(img, prompts):
+    image, img_prompts = img['image'], img['points']
+    point_prompts = []
+    box_prompts = []
+    for prompt in img_prompts:
+        for p in range(len(prompt)):
+            prompt[p] = int(prompt[p])
+        if prompt[2] == 2 and prompt[5] == 3:  # box prompt
+            if len(box_prompts) != 0:
+                raise gr.Error("Please input only one BBox.", duration=3)
+            box_prompts.append([prompt[0], prompt[1], prompt[3], prompt[4]])
+        elif prompt[2] == 1 and prompt[5] == 4:  # Positive point prompt
+            point_prompts.append((1, (prompt[0], prompt[1])))
+        elif prompt[2] == 0 and prompt[5] == 4:  # Negative point prompt
+            point_prompts.append((0, (prompt[0], prompt[1])))
+
+    if "scribble" in prompts:
+        del prompts["scribble"]
+
+    if len(point_prompts) > 0:
+        prompts['point'] = point_prompts
+    elif 'point' in prompts:
+        del prompts['point']
+
+    if len(box_prompts) > 0:
+        prompts['bbox'] = box_prompts
+    elif 'bbox' in prompts:
+        del prompts['bbox']
+
+    zim_mask, sam_mask = run_model(image, prompts)
+
+    return image, zim_mask, sam_mask, prompts
+
+def get_examples():
+    assets_dir = os.path.join(os.path.dirname(__file__), 'examples')
+    images = os.listdir(assets_dir)
+    return [os.path.join(assets_dir, img) for img in images]
+
+if __name__ == "__main__":
+    backbone = "vit_b"
+    
+    # load ZIM
+    ckpt_mat = "ckpts/zim_vit_b_2043"
+    zim = zim_model_registry[backbone](checkpoint=ckpt_mat)
+    if torch.cuda.is_available():
+        zim.cuda()
+    zim_predictor = ZimPredictor(zim)
+    zim_mask_generator = ZimAutomaticMaskGenerator(
+        zim, 
+        pred_iou_thresh=0.7, 
+        points_per_batch=8,
+        stability_score_thresh=0.9, 
+    )
+
+    # load SAM
+    ckpt_sam = "ckpts/sam_vit_b_01ec64.pth"
+    sam = sam_model_registry[backbone](checkpoint=ckpt_sam)
+    if torch.cuda.is_available():
+        sam.cuda()
+    sam_predictor = SamPredictor(sam)
+    sam_mask_generator = SamAutomaticMaskGenerator(
+        sam,
+        points_per_batch=8,
+    )
+    
+    with gr.Blocks() as demo:
+        gr.Markdown("# <center> [Demo] ZIM: Zero-Shot Image Matting for Anything")
+
+        prompts = gr.State(dict())
+        img = gr.Image(visible=False)
+        example_image = gr.Image(visible=False)
+        
+        with gr.Row():
+            with gr.Column():
+                # Point and Bbox prompt
+                with gr.Tab(label="Point or Box"):
+                    img_with_point_or_box = ImagePrompter(
+                        label="query image", 
+                        sources="upload"
+                    )
+                    interactions = "Left Click (Pos) | Middle/Right Click (Neg) | Press Move (Box)"
+                    gr.Markdown("<h3 style='text-align: center'> {} </h3>".format(interactions))
+                    run_bttn = gr.Button("Run")
+                    amg_bttn = gr.Button("Automatic Mask Generation")
+
+                # Scribble prompt
+                with gr.Tab(label="Scribble"):
+                    img_with_scribble = gr.ImageEditor(
+                        label="Scribble", 
+                        brush=gr.Brush(colors=["#00FF00"], default_size=15),
+                        sources="upload", 
+                        transforms=None, 
+                        layers=False
+                    )
+                    interactions = "Press Move (Scribble)"
+                    gr.Markdown("<h3 style='text-align: center'> Step 1. Select Draw button </h3>")
+                    gr.Markdown("<h3 style='text-align: center'> Step 2. {} </h3>".format(interactions))
+                    scribble_bttn = gr.Button("Run")
+                    scribble_reset_bttn = gr.Button("Reset Scribbles")
+                    amg_scribble_bttn = gr.Button("Automatic Mask Generation")
+                
+                # Example image
+                gr.Examples(get_examples(), inputs=[example_image])
+
+            # with gr.Row():
+            with gr.Column():
+                with gr.Tab(label="ZIM Image"):
+                    img_with_zim_mask = gr.Image(
+                        label="ZIM Image", 
+                        interactive=False
+                    )
+
+                with gr.Tab(label="ZIM Mask"):
+                    zim_mask = gr.Image(
+                        label="ZIM Mask", 
+                        image_mode="L", 
+                        interactive=False
+                    )
+                with gr.Tab(label="ZIM Auto Mask"):
+                    zim_amg = gr.Image(
+                        label="ZIM Auto Mask", 
+                        interactive=False
+                    )
+                    
+            with gr.Column():
+                with gr.Tab(label="SAM Image"):
+                    img_with_sam_mask = gr.Image(
+                        label="SAM image", 
+                        interactive=False
+                    )
+
+                with gr.Tab(label="SAM Mask"):
+                    sam_mask = gr.Image(
+                        label="SAM Mask", 
+                        image_mode="L", 
+                        interactive=False
+                    )
+                    
+                with gr.Tab(label="SAM Auto Mask"):
+                    sam_amg = gr.Image(
+                        label="SAM Auto Mask", 
+                        interactive=False
+                    )
+                    
+        example_image.change(
+            reset_example_image,
+            [example_image, prompts],
+            [
+                img,
+                img_with_point_or_box,
+                img_with_scribble,
+                img_with_zim_mask,
+                img_with_sam_mask,
+                zim_amg,
+                sam_amg,
+                zim_mask,
+                sam_mask,
+                prompts,
+            ]
+        )
+
+        img_with_point_or_box.upload(
+            reset_image,
+            [img_with_point_or_box, prompts],
+            [
+                img,
+                img_with_scribble,
+                img_with_zim_mask,
+                img_with_sam_mask,
+                zim_amg,
+                sam_amg,
+                zim_mask,
+                sam_mask,
+                prompts,
+            ],
+        )
+
+        amg_bttn.click(
+            run_amg,
+            [img],
+            [zim_amg, sam_amg]
+        )
+        amg_scribble_bttn.click(
+            run_amg,
+            [img],
+            [zim_amg, sam_amg]
+        )
+        
+        run_bttn.click(
+            get_point_or_box_prompts,
+            [img_with_point_or_box, prompts],
+            [img, zim_mask, sam_mask, prompts]
+        )
+
+        zim_mask.change(
+            draw_images,
+            [img, zim_mask, prompts],
+            [
+                img, img_with_zim_mask, 
+            ],
+        )
+        sam_mask.change(
+            draw_images,
+            [img, sam_mask, prompts],
+            [
+                img, img_with_sam_mask, 
+            ],
+        )
+        
+        scribble_reset_bttn.click(
+            reset_scribble,
+            [img, img_with_scribble, prompts],
+            [img_with_scribble, zim_mask, sam_mask],
+        )
+        scribble_bttn.click(
+            update_scribble,
+            [img, img_with_scribble, prompts],
+            [zim_mask, sam_mask, prompts],
+        )
+
+    demo.queue()
+    demo.launch()

ckpts/sam_vit_b_01ec64.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec2df62732614e57411cdcf32a23ffdf28910380d03139ee0f4fcbe91eb8c912
+size 375042383

ckpts/zim_vit_b_2043/decoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a99d14bac3dc28b17809020b1711b1991bd5b7eb18f678c3624ce508748cfa11
+size 19330176

ckpts/zim_vit_b_2043/encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6123ea10722dd43a0f8bbb96cd6a0532d8a65e2b1e8339f0cd6593d494807809
+size 360680416

ckpts/zim_vit_l_2092/decoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c24a42d79053d20594760ad9afdef9cca985f29493f3e1e0fe3462ca291b57f7
+size 19330176

ckpts/zim_vit_l_2092/encoder.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b0360b3c32dfa11555fb1de27cd3533cec1e932204da0dec2b3772045dc7db3
+size 1235692110

config/__init__.py

@@ -0,0 +1 @@
+from config.config import generate_config

config/config.py

@@ -0,0 +1,66 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+from easydict import EasyDict as edict
+
+config_ = edict()
+
+"""
+    Common configs
+"""
+config_.data_root = "/mnt/tmp"
+config_.use_ddp = True
+config_.use_amp = False
+config_.local_rank = 0
+config_.world_size = 1
+config_.random_seed = 3407
+"""
+    Network configs
+"""
+config_.network = edict()
+config_.network.encoder = "vit_b"
+config_.network.decoder = "zim"
+config_.network.encode_kernel = 21
+"""
+    Evaluation configs
+"""
+config_.eval = edict()
+config_.eval.workers = 4
+config_.eval.image_size = 1024
+config_.eval.prompt_type = "point,bbox"
+config_.eval.model_list = "zim,sam"
+config_.eval.zim_weights = ""
+config_.eval.sam_weights = ""
+"""
+    Dataset configs
+"""
+config_.dataset = edict()
+config_.dataset.valset = "MicroMat3K"
+config_.dataset.data_type = "fine,coarse"
+config_.dataset.data_list_txt = "data_list.txt"
+
+
+def remove_prefix(text, prefix):
+    if text.startswith(prefix):
+        return text[len(prefix) :]
+    return text
+
+
+def generate_config(args):
+    # merge args & config
+    for k, v in args.items():
+        if k.startswith("network_"):
+            config_["network"][remove_prefix(k, "network_")] = v
+        elif k.startswith("eval_"):
+            config_["eval"][remove_prefix(k, "eval_")] = v
+        elif k.startswith("dataset_"):
+            config_["dataset"][remove_prefix(k, "dataset_")] = v
+        elif k == "amp":
+            config_["use_amp"] = v
+        else:
+            config_[k] = v
+    return config_

packages.txt

@@ -0,0 +1 @@
+wget

pre-requirements.txt

@@ -0,0 +1,3 @@
+torch
+torchvision
+numpy==1.24.4

requirements.txt

@@ -0,0 +1,7 @@
+easydict
+opencv-python
+gradio==4.38.1
+gradio-image-prompter
+fastapi==0.112.2
+git+https://github.com/facebookresearch/segment-anything.git
+onnxruntime-gpu==1.17.0

zim/__init__.py

@@ -0,0 +1,9 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from .build_model import build_zim_model, zim_model_registry
+from .predictor import ZimPredictor
+from .automatic_mask_generator import ZimAutomaticMaskGenerator

zim/automatic_mask_generator.py

@@ -0,0 +1,378 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import numpy as np
+import torch
+from torchvision.ops.boxes import batched_nms, box_area
+
+from typing import Any, Dict, List, Optional, Tuple
+
+from .modeling.zim import Zim
+from .predictor import ZimPredictor
+from .utils.amg import (
+    MaskData,
+    area_from_rle,
+    batch_iterator,
+    batched_mask_to_box,
+    box_xyxy_to_xywh,
+    build_all_layer_point_grids,
+    calculate_stability_score,
+    coco_encode_rle,
+    generate_crop_boxes,
+    is_box_near_crop_edge,
+    mask_to_rle_pytorch,
+    remove_small_regions,
+    rle_to_mask,
+    uncrop_boxes_xyxy,
+    uncrop_masks,
+    uncrop_points,
+)
+
+
+class ZimAutomaticMaskGenerator:
+    def __init__(
+        self,
+        model: Zim,
+        points_per_side: Optional[int] = 32,
+        points_per_batch: int = 64,
+        pred_iou_thresh: float = 0.88,
+        stability_score_thresh: float = 0.9,
+        stability_score_offset: float = 0.1,
+        box_nms_thresh: float = 0.7,
+        crop_n_layers: int = 0,
+        crop_nms_thresh: float = 0.7,
+        crop_overlap_ratio: float = 512 / 1500,
+        crop_n_points_downscale_factor: int = 1,
+        point_grids: Optional[List[np.ndarray]] = None,
+        min_mask_region_area: int = 0,
+        output_mode: str = "binary_mask",
+    ) -> None:
+        """
+        Using a SAM model, generates masks for the entire image.
+        Generates a grid of point prompts over the image, then filters
+        low quality and duplicate masks. The default settings are chosen
+        for SAM with a ViT-H backbone.
+
+        Arguments:
+          model (Sam): The SAM model to use for mask prediction.
+          points_per_side (int or None): The number of points to be sampled
+            along one side of the image. The total number of points is
+            points_per_side**2. If None, 'point_grids' must provide explicit
+            point sampling.
+          points_per_batch (int): Sets the number of points run simultaneously
+            by the model. Higher numbers may be faster but use more GPU memory.
+          pred_iou_thresh (float): A filtering threshold in [0,1], using the
+            model's predicted mask quality.
+          stability_score_thresh (float): A filtering threshold in [0,1], using
+            the stability of the mask under changes to the cutoff used to binarize
+            the model's mask predictions.
+          stability_score_offset (float): The amount to shift the cutoff when
+            calculated the stability score.
+          box_nms_thresh (float): The box IoU cutoff used by non-maximal
+            suppression to filter duplicate masks.
+          crop_n_layers (int): If >0, mask prediction will be run again on
+            crops of the image. Sets the number of layers to run, where each
+            layer has 2**i_layer number of image crops.
+          crop_nms_thresh (float): The box IoU cutoff used by non-maximal
+            suppression to filter duplicate masks between different crops.
+          crop_overlap_ratio (float): Sets the degree to which crops overlap.
+            In the first crop layer, crops will overlap by this fraction of
+            the image length. Later layers with more crops scale down this overlap.
+          crop_n_points_downscale_factor (int): The number of points-per-side
+            sampled in layer n is scaled down by crop_n_points_downscale_factor**n.
+          point_grids (list(np.ndarray) or None): A list over explicit grids
+            of points used for sampling, normalized to [0,1]. The nth grid in the
+            list is used in the nth crop layer. Exclusive with points_per_side.
+          min_mask_region_area (int): If >0, postprocessing will be applied
+            to remove disconnected regions and holes in masks with area smaller
+            than min_mask_region_area. Requires opencv.
+          output_mode (str): The form masks are returned in. Can be 'binary_mask',
+            'uncompressed_rle', or 'coco_rle'. 'coco_rle' requires pycocotools.
+            For large resolutions, 'binary_mask' may consume large amounts of
+            memory.
+        """
+
+        assert (points_per_side is None) != (
+            point_grids is None
+        ), "Exactly one of points_per_side or point_grid must be provided."
+        if points_per_side is not None:
+            self.point_grids = build_all_layer_point_grids(
+                points_per_side,
+                crop_n_layers,
+                crop_n_points_downscale_factor,
+            )
+        elif point_grids is not None:
+            self.point_grids = point_grids
+        else:
+            raise ValueError("Can't have both points_per_side and point_grid be None.")
+
+        assert output_mode in [
+            "binary_mask",
+            "uncompressed_rle",
+            "coco_rle",
+        ], f"Unknown output_mode {output_mode}."
+        if output_mode == "coco_rle":
+            from pycocotools import mask as mask_utils  # type: ignore # noqa: F401
+
+        if min_mask_region_area > 0:
+            import cv2  # type: ignore # noqa: F401
+
+        self.predictor = ZimPredictor(model)
+        self.points_per_batch = points_per_batch
+        self.pred_iou_thresh = pred_iou_thresh
+        self.stability_score_thresh = stability_score_thresh
+        self.stability_score_offset = stability_score_offset
+        self.box_nms_thresh = box_nms_thresh
+        self.crop_n_layers = crop_n_layers
+        self.crop_nms_thresh = crop_nms_thresh
+        self.crop_overlap_ratio = crop_overlap_ratio
+        self.crop_n_points_downscale_factor = crop_n_points_downscale_factor
+        self.min_mask_region_area = min_mask_region_area
+        self.output_mode = output_mode
+
+    @torch.no_grad()
+    def generate(self, image: np.ndarray) -> List[Dict[str, Any]]:
+        """
+        Generates masks for the given image.
+
+        Arguments:
+          image (np.ndarray): The image to generate masks for, in HWC uint8 format.
+
+        Returns:
+           list(dict(str, any)): A list over records for masks. Each record is
+             a dict containing the following keys:
+               segmentation (dict(str, any) or np.ndarray): The mask. If
+                 output_mode='binary_mask', is an array of shape HW. Otherwise,
+                 is a dictionary containing the RLE.
+               bbox (list(float)): The box around the mask, in XYWH format.
+               area (int): The area in pixels of the mask.
+               predicted_iou (float): The model's own prediction of the mask's
+                 quality. This is filtered by the pred_iou_thresh parameter.
+               point_coords (list(list(float))): The point coordinates input
+                 to the model to generate this mask.
+               stability_score (float): A measure of the mask's quality. This
+                 is filtered on using the stability_score_thresh parameter.
+               crop_box (list(float)): The crop of the image used to generate
+                 the mask, given in XYWH format.
+        """
+
+        # Generate masks
+        mask_data = self._generate_masks(image)
+
+        # Filter small disconnected regions and holes in masks
+        if self.min_mask_region_area > 0:
+            mask_data = self.postprocess_small_regions(
+                mask_data,
+                self.min_mask_region_area,
+                max(self.box_nms_thresh, self.crop_nms_thresh),
+            )
+
+        # Encode masks
+        if self.output_mode == "coco_rle":
+            mask_data["segmentations"] = [coco_encode_rle(rle) for rle in mask_data["rles"]]
+        elif self.output_mode == "binary_mask":
+            mask_data["segmentations"] = [rle_to_mask(rle) for rle in mask_data["rles"]]
+        else:
+            mask_data["segmentations"] = mask_data["rles"]
+
+        # Write mask records
+        curr_anns = []
+        for idx in range(len(mask_data["segmentations"])):
+            ann = {
+                "segmentation": mask_data["segmentations"][idx],
+                "logit": mask_data["logits"][idx],
+                "area": area_from_rle(mask_data["rles"][idx]),
+                "bbox": box_xyxy_to_xywh(mask_data["boxes"][idx]).tolist(),
+                "predicted_iou": mask_data["iou_preds"][idx].item(),
+                "point_coords": [mask_data["points"][idx].tolist()],
+                "stability_score": mask_data["stability_score"][idx].item(),
+                "crop_box": box_xyxy_to_xywh(mask_data["crop_boxes"][idx]).tolist(),
+            }
+            curr_anns.append(ann)
+
+        return curr_anns
+
+    def _generate_masks(self, image: np.ndarray) -> MaskData:
+        orig_size = image.shape[:2]
+        crop_boxes, layer_idxs = generate_crop_boxes(
+            orig_size, self.crop_n_layers, self.crop_overlap_ratio
+        )
+
+        # Iterate over image crops
+        data = MaskData()
+        for crop_box, layer_idx in zip(crop_boxes, layer_idxs):
+            crop_data = self._process_crop(image, crop_box, layer_idx, orig_size)
+            data.cat(crop_data)
+
+        # Remove duplicate masks between crops
+        if len(crop_boxes) > 1:
+            # Prefer masks from smaller crops
+            scores = 1 / box_area(data["crop_boxes"])
+            scores = scores.to(data["boxes"].device)
+            keep_by_nms = batched_nms(
+                data["boxes"].float(),
+                scores,
+                torch.zeros_like(data["boxes"][:, 0]),  # categories
+                iou_threshold=self.crop_nms_thresh,
+            )
+            data.filter(keep_by_nms)
+
+        data.to_numpy()
+        return data
+
+    def _process_crop(
+        self,
+        image: np.ndarray,
+        crop_box: List[int],
+        crop_layer_idx: int,
+        orig_size: Tuple[int, ...],
+    ) -> MaskData:
+        # Crop the image and calculate embeddings
+        x0, y0, x1, y1 = crop_box
+        cropped_im = image[y0:y1, x0:x1, :]
+        cropped_im_size = cropped_im.shape[:2]
+        self.predictor.set_image(cropped_im)
+
+        # Get points for this crop
+        points_scale = np.array(cropped_im_size)[None, ::-1]
+        points_for_image = self.point_grids[crop_layer_idx] * points_scale
+
+        # Generate masks for this crop in batches
+        data = MaskData()
+        for (points,) in batch_iterator(self.points_per_batch, points_for_image):
+            batch_data = self._process_batch(points, cropped_im_size, crop_box, orig_size)
+            data.cat(batch_data)
+            del batch_data
+        self.predictor.reset_image()
+
+        # Remove duplicates within this crop.
+        keep_by_nms = batched_nms(
+            data["boxes"].float(),
+            data["iou_preds"],
+            torch.zeros_like(data["boxes"][:, 0]),  # categories
+            iou_threshold=self.box_nms_thresh,
+        )
+        data.filter(keep_by_nms)
+
+        # Return to the original image frame
+        data["boxes"] = uncrop_boxes_xyxy(data["boxes"], crop_box)
+        data["points"] = uncrop_points(data["points"], crop_box)
+        data["crop_boxes"] = torch.tensor([crop_box for _ in range(len(data["rles"]))])
+
+        return data
+
+    def _process_batch(
+        self,
+        points: np.ndarray,
+        im_size: Tuple[int, ...],
+        crop_box: List[int],
+        orig_size: Tuple[int, ...],
+    ) -> MaskData:
+        orig_h, orig_w = orig_size
+
+        # Run model on this batch
+        transformed_points = self.predictor.transform.apply_coords(points, im_size)
+        in_points = torch.as_tensor(transformed_points, device=self.predictor.device)
+        in_labels = torch.ones(in_points.shape[0], dtype=torch.int, device=in_points.device)
+        masks, iou_preds, _ = self.predictor.predict_torch(
+            in_points[:, None, :],
+            in_labels[:, None],
+            multimask_output=True,
+            return_logits=True,
+        )
+
+        # Serialize predictions and store in MaskData
+        data = MaskData(
+            masks=masks.flatten(0, 1),
+            logits=(masks.flatten(0, 1) * 255).byte(),
+            iou_preds=iou_preds.flatten(0, 1),
+            points=torch.as_tensor(points.repeat(masks.shape[1], axis=0)),
+        )
+        del masks
+
+        # Filter by predicted IoU
+        if self.pred_iou_thresh > 0.0:
+            keep_mask = data["iou_preds"] > self.pred_iou_thresh
+            data.filter(keep_mask)
+
+        # Calculate stability score
+        data["stability_score"] = calculate_stability_score(
+            data["masks"], self.predictor.model.mask_threshold, self.stability_score_offset
+        )
+        if self.stability_score_thresh > 0.0:
+            keep_mask = data["stability_score"] >= self.stability_score_thresh
+            data.filter(keep_mask)
+
+        # Threshold masks and calculate boxes
+        data["masks"] = data["masks"] > self.predictor.model.mask_threshold
+        data["boxes"] = batched_mask_to_box(data["masks"])
+
+        # Filter boxes that touch crop boundaries
+        keep_mask = ~is_box_near_crop_edge(data["boxes"], crop_box, [0, 0, orig_w, orig_h])
+        if not torch.all(keep_mask):
+            data.filter(keep_mask)
+
+        # Compress to RLE
+        data["masks"] = uncrop_masks(data["masks"], crop_box, orig_h, orig_w)
+        data["logits"] = uncrop_masks(data["logits"], crop_box, orig_h, orig_w)
+        data["rles"] = mask_to_rle_pytorch(data["masks"])
+        del data["masks"]
+
+        return data
+
+    @staticmethod
+    def postprocess_small_regions(
+        mask_data: MaskData, min_area: int, nms_thresh: float
+    ) -> MaskData:
+        """
+        Removes small disconnected regions and holes in masks, then reruns
+        box NMS to remove any new duplicates.
+
+        Edits mask_data in place.
+
+        Requires open-cv as a dependency.
+        """
+        if len(mask_data["rles"]) == 0:
+            return mask_data
+
+        # Filter small disconnected regions and holes
+        new_masks = []
+        scores = []
+        for rle in mask_data["rles"]:
+            mask = rle_to_mask(rle)
+
+            mask, changed = remove_small_regions(mask, min_area, mode="holes")
+            unchanged = not changed
+            mask, changed = remove_small_regions(mask, min_area, mode="islands")
+            unchanged = unchanged and not changed
+
+            new_masks.append(torch.as_tensor(mask).unsqueeze(0))
+            # Give score=0 to changed masks and score=1 to unchanged masks
+            # so NMS will prefer ones that didn't need postprocessing
+            scores.append(float(unchanged))
+
+        # Recalculate boxes and remove any new duplicates
+        masks = torch.cat(new_masks, dim=0)
+        boxes = batched_mask_to_box(masks)
+        keep_by_nms = batched_nms(
+            boxes.float(),
+            torch.as_tensor(scores),
+            torch.zeros_like(boxes[:, 0]),  # categories
+            iou_threshold=nms_thresh,
+        )
+
+        # Only recalculate RLEs for masks that have changed
+        for i_mask in keep_by_nms:
+            if scores[i_mask] == 0.0:
+                mask_torch = masks[i_mask].unsqueeze(0)
+                mask_data["rles"][i_mask] = mask_to_rle_pytorch(mask_torch)[0]
+                mask_data["boxes"][i_mask] = boxes[i_mask]  # update res directly
+        mask_data.filter(keep_by_nms)
+
+        return mask_data

zim/build_model.py

@@ -0,0 +1,29 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+import os
+import torch
+
+from .modeling.zim import Zim
+from .modeling.encoder import ZIM_Encoder
+from .modeling.decoder import ZIM_Decoder
+
+def build_zim_model(checkpoint):
+    
+    encoder = ZIM_Encoder(os.path.join(checkpoint, "encoder.onnx"))
+    decoder = ZIM_Decoder(os.path.join(checkpoint, "decoder.onnx"))
+    net = Zim(encoder, decoder)
+
+    return net
+
+zim_model_registry = {
+    "default": build_zim_model,
+    "vit_l": build_zim_model,
+    "vit_b": build_zim_model,
+}
+

zim/modeling/decoder.py

@@ -0,0 +1,90 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+import torch
+from typing import Any, Callable
+import onnxruntime
+import numpy as np
+
+def np2tensor(np_array, device):
+    return torch.from_numpy(np_array).to(device)
+
+def tensor2np(torch_tensor):
+    if torch_tensor is None:
+        return None
+    
+    return torch_tensor.detach().cpu().numpy()
+    
+class ZIM_Decoder():
+    def __init__(self, onnx_path, num_threads=16):
+        self.onnx_path = onnx_path
+        
+        sessionOptions = onnxruntime.SessionOptions()
+        sessionOptions.intra_op_num_threads = num_threads
+        sessionOptions.inter_op_num_threads = num_threads
+        providers = ["CPUExecutionProvider"]
+
+        self.ort_session = onnxruntime.InferenceSession(
+            onnx_path, sess_options=sessionOptions, providers=providers
+        )
+        self.num_mask_tokens = 4
+        
+    def cuda(self, device_id=0):
+        providers = [
+            (
+                "CUDAExecutionProvider",
+                {
+                    "device_id": device_id,
+                },
+            ),
+        ]
+
+        self.ort_session.set_providers(providers)
+        
+    def forward(
+        self, 
+        interm_feats,
+        image_embeddings, 
+        points, 
+        boxes,
+        attn_mask,
+    ):
+        device = image_embeddings.device
+        
+        ort_inputs = {
+            "feat_D0": tensor2np(interm_feats[0]),
+            "feat_D1": tensor2np(interm_feats[1]),
+            "feat_D2": tensor2np(interm_feats[2]),
+            "image_embeddings": tensor2np(image_embeddings),
+            "attn_mask": tensor2np(attn_mask),
+        }
+        
+        if points is not None:
+            point_coords, point_labels = points
+            ort_inputs["point_coords"] = tensor2np(point_coords.float())
+            ort_inputs["point_labels"] = tensor2np(point_labels.float())
+            
+            # add paddings as done in SAM
+            padding_point = np.zeros((ort_inputs["point_coords"].shape[0], 1, 2), dtype=np.float32) - 0.5
+            padding_label = -np.ones((ort_inputs["point_labels"].shape[0], 1), dtype=np.float32)
+            ort_inputs["point_coords"] = np.concatenate([ort_inputs["point_coords"], padding_point], axis=1)
+            ort_inputs["point_labels"] = np.concatenate([ort_inputs["point_labels"], padding_label], axis=1)
+            
+        if boxes is not None:
+            ort_inputs["point_coords"] = tensor2np(boxes.reshape(-1, 2, 2))
+            ort_inputs["point_labels"] = np.array([[2, 3]], dtype=np.float32).repeat(boxes.shape[0], 0)
+        
+        masks, iou_predictions = self.ort_session.run(None, ort_inputs)
+        
+        masks = np2tensor(masks, device)
+        iou_predictions = np2tensor(iou_predictions, device)
+        
+        return masks, iou_predictions
+    
+    __call__: Callable[..., Any] = forward
+    

zim/modeling/encoder.py

@@ -0,0 +1,62 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+import torch
+from typing import Any, Callable
+import onnxruntime
+
+def np2tensor(np_array, device):
+    return torch.from_numpy(np_array).to(device)
+
+def tensor2np(torch_tensor):
+    return torch_tensor.detach().cpu().numpy()
+    
+class ZIM_Encoder():
+    def __init__(self, onnx_path, num_threads=16):
+        self.onnx_path = onnx_path
+        
+        sessionOptions = onnxruntime.SessionOptions()
+        sessionOptions.intra_op_num_threads = num_threads
+        sessionOptions.inter_op_num_threads = num_threads
+        providers = ["CPUExecutionProvider"]
+
+        self.ort_session = onnxruntime.InferenceSession(
+            onnx_path, sess_options=sessionOptions, providers=providers
+        )
+        
+    def cuda(self, device_id=0):
+        providers = [
+            (
+                "CUDAExecutionProvider",
+                {
+                    "device_id": device_id,
+                },
+            ),
+        ]
+
+        self.ort_session.set_providers(providers)
+        
+    def forward(
+        self, 
+        image, 
+    ):
+        device = image.device
+        
+        ort_inputs = {
+            "image": tensor2np(image),
+        }
+        image_embeddings, feat_D0, feat_D1, feat_D2 = self.ort_session.run(None, ort_inputs)
+        
+        image_embeddings = np2tensor(image_embeddings, device)
+        feat_D0 = np2tensor(feat_D0, device)
+        feat_D1 = np2tensor(feat_D1, device)
+        feat_D2 = np2tensor(feat_D2, device)
+        
+        return image_embeddings, (feat_D0, feat_D1, feat_D2)
+    
+    __call__: Callable[..., Any] = forward

zim/modeling/zim.py

@@ -0,0 +1,190 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import torch
+from torch import nn
+from torch.nn import functional as F
+from typing import Any, Dict, List
+
+def gaussian(sigma=6):
+    """
+    2D Gaussian Kernel Generation.
+    """
+    size = 6 * sigma + 3
+    x = torch.arange(0, size, 1)
+    y = x[:, None]
+    x0, y0 = 3 * sigma + 1, 3 * sigma + 1
+    g = torch.exp(- ((x - x0) ** 2 + (y - y0) ** 2) / (2 * sigma ** 2))
+    return g
+
+class Zim(nn.Module):
+    def __init__(
+        self,
+        encoder,
+        decoder,
+        *,
+        image_size: int = 1024,
+        pixel_mean: List[float] = [123.675, 116.28, 103.53],
+        pixel_std: List[float] = [58.395, 57.12, 57.375],
+    ) -> None:
+        """
+        SAM predicts object masks from an image and input prompts.
+
+        Arguments:
+          encoder : The backbone used to encode the
+            image into image embeddings that allow for efficient mask prediction.
+          decoder : Predicts masks from the image embeddings and given prompts.
+          pixel_mean (list(float)): Mean values for normalizing pixels in the input image.
+          pixel_std (list(float)): Std values for normalizing pixels in the input image.
+        """
+        super().__init__()
+        self.encoder = encoder
+        self.decoder = decoder
+        self.output_activation = nn.Sigmoid()
+
+        self.image_size = image_size
+        self.register_buffer(
+            "pixel_mean", torch.Tensor(pixel_mean).view(-1, 1, 1), False
+        )
+        self.register_buffer("pixel_std", torch.Tensor(pixel_std).view(-1, 1, 1), False)
+
+        self.mask_threshold: float = 0.5
+        self.image_format: str = "RGB"
+        self.num_mask_tokens = decoder.num_mask_tokens
+        
+        self.encode_stride = 16
+        self.encode_kernel = 21
+        self.attn_mask_size = 64
+        self.g = gaussian(self.encode_kernel)
+        
+        self.output_conv = nn.Conv2d(
+            self.num_mask_tokens, 
+            self.num_mask_tokens, 
+            kernel_size=1, stride=1, padding=0,
+        )
+
+    @property
+    def device(self) -> Any:
+        return self.pixel_mean.device
+    
+    def cuda(self, device_id=None):
+        if type(device_id) == torch.device:
+            device_id = device_id.index
+            
+        if device_id is None:
+            device_id = 0
+        
+        device = torch.device(f"cuda:{device_id}")
+        super(Zim, self).cuda(device)
+        
+        self.encoder.cuda(device_id)
+        self.decoder.cuda(device_id)
+        
+        return self
+
+    def postprocess_masks(
+        self, masks: torch.Tensor, input_size: List[int], original_size: torch.Tensor
+    ) -> torch.Tensor:
+        """
+        Remove padding and upscale masks to the original image size.
+
+        Arguments:
+          masks (torch.Tensor): Batched masks from the decoder,
+            in BxCxHxW format.
+          input_size (tuple(int, int)): The size of the image input to the
+            model, in (H, W) format. Used to remove padding.
+          original_size (tuple(int, int)): The original size of the image
+            before resizing for input to the model, in (H, W) format.
+
+        Returns:
+          (torch.Tensor): Batched masks in BxCxHxW format, where (H, W)
+            is given by original_size.
+        """
+        masks = F.interpolate(
+            masks,
+            (self.image_size, self.image_size),
+            mode="bilinear",
+            align_corners=False,
+        )
+        masks = masks[..., : input_size[0], : input_size[1]]
+        masks = F.interpolate(
+            masks, original_size, mode="bilinear", align_corners=False
+        )
+        return masks
+
+    def preprocess(self, x: torch.Tensor) -> torch.Tensor:
+        """Normalize pixel values and pad to a square input."""
+        # Normalize colors
+        x = (x - self.pixel_mean) / self.pixel_std
+
+        # Pad
+        h, w = x.shape[-2:]
+        padh = self.image_size - h
+        padw = self.image_size - w
+        x = F.pad(x, (0, padw, 0, padh))
+        return x
+
+    def bbox_attn_mask(self, boxes):
+        """Prompt-aware Masked Attention: box prompt (binary attn mask) """
+        bs = boxes.shape[0]
+        attn_mask = torch.zeros((bs, self.attn_mask_size, self.attn_mask_size), device=boxes.device)
+            
+        # attn_weight = attn_weight.masked_fill(m.logical_not(), -1e4)
+        
+        for n in range(bs):
+            xmin, ymin, xmax, ymax = boxes[n]
+            
+            xmin, xmax  = min(xmin, xmax), max(xmin, xmax)
+            ymin, ymax  = min(ymin, ymax), max(ymin, ymax)
+            
+            xmin, xmax = int(xmin / self.encode_stride), int(xmax / self.encode_stride)
+            ymin, ymax = int(ymin / self.encode_stride), int(ymax / self.encode_stride)
+            
+            xmin, ymin = max(0, xmin), max(0, ymin)
+            xmax = min(self.attn_mask_size, xmax+1)
+            ymax = min(self.attn_mask_size, ymax+1)
+            
+            attn_mask[n, ymin:ymax, xmin:xmax] = 1
+            
+        return attn_mask
+    
+    def point_attn_mask(self, point_coords):
+        """Prompt-aware Masked Attention: point prompt (soft attn mask) """
+        bs = point_coords.shape[0]
+        attn_mask = torch.zeros((bs, self.attn_mask_size, self.attn_mask_size), device=point_coords.device)
+        
+        if self.g.device != point_coords.device:
+            self.g = self.g.to(point_coords.device)
+                
+        for n in range(bs):
+            for point in point_coords[n]:
+                x, y = int(point[0] / self.encode_stride), int(point[1].item() / self.encode_stride)
+                
+                # outside image boundary
+                if x < 0 or y < 0 or x >= self.attn_mask_size or y >= self.attn_mask_size:
+                    continue
+
+                # upper left
+                ul = int(round(x - 3 * self.encode_kernel - 1)), int(round(y - 3 * self.encode_kernel - 1))
+                # bottom right
+                br = int(round(x + 3 * self.encode_kernel + 2)), int(round(y + 3 * self.encode_kernel + 2))
+
+                c, d = int(max(0, -ul[0])), int(min(br[0], self.attn_mask_size) - ul[0])
+                a, b = int(max(0, -ul[1])), int(min(br[1], self.attn_mask_size) - ul[1])
+
+                cc, dd = int(max(0, ul[0])), int(min(br[0], self.attn_mask_size))
+                aa, bb = int(max(0, ul[1])), int(min(br[1], self.attn_mask_size))
+
+                attn_mask[n, aa:bb, cc:dd] = torch.maximum(
+                    attn_mask[n, aa:bb, cc:dd], self.g[a:b, c:d]
+                )
+                
+        return attn_mask
+    
+    

zim/predictor.py

@@ -0,0 +1,275 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import numpy as np
+import torch
+from torch.nn import functional as F
+from torch.nn.parallel import DistributedDataParallel as DDP
+from typing import Optional, Tuple, List
+
+from .utils import ResizeLongestSide
+
+class ZimPredictor:
+    def __init__(
+        self,
+        model,
+    ) -> None:
+        """
+        Uses SAM to calculate the image embedding for an image, and then
+        allow repeated, efficient mask prediction given prompts.
+
+        Arguments:
+          sam_model (Sam): The model to use for mask prediction.
+        """
+        super().__init__()
+        self.model = model.module if isinstance(model, DDP) else model
+        self.transform = ResizeLongestSide(self.model.image_size)
+        self.reset_image()
+
+    def set_image(
+        self,
+        image: np.ndarray,
+        image_format: str = "RGB",
+    ) -> None:
+        """
+        Calculates the image embeddings for the provided image, allowing
+        masks to be predicted with the 'predict' method.
+
+        Arguments:
+          image (np.ndarray): The image for calculating masks. Expects an
+            image in HWC uint8 format, with pixel values in [0, 255].
+          image_format (str): The color format of the image, in ['RGB', 'BGR'].
+        """
+        assert image_format in [
+            "RGB",
+            "BGR",
+        ], f"image_format must be in ['RGB', 'BGR'], is {image_format}."
+        if image_format != self.model.image_format:
+            image = image[..., ::-1]
+
+        # Transform the image to the form expected by the model
+        input_image = self.transform.apply_image(image)
+        input_image_torch = torch.as_tensor(input_image, device=self.device)
+        input_image_torch = input_image_torch.permute(2, 0, 1).contiguous()[None, :, :, :]
+
+        self.set_torch_image(input_image_torch, image.shape[:2])
+
+    @torch.no_grad()
+    def set_torch_image(
+        self,
+        transformed_image: torch.Tensor,
+        original_image_size: Tuple[int, ...],
+    ) -> None:
+        """
+        Calculates the image embeddings for the provided image, allowing
+        masks to be predicted with the 'predict' method. Expects the input
+        image to be already transformed to the format expected by the model.
+
+        Arguments:
+          transformed_image (torch.Tensor): The input image, with shape
+            1x3xHxW, which has been transformed with ResizeLongestSide.
+          original_image_size (tuple(int, int)): The size of the image
+            before transformation, in (H, W) format.
+        """
+        assert (
+            len(transformed_image.shape) == 4
+            and transformed_image.shape[1] == 3
+            and max(*transformed_image.shape[2:]) == self.model.image_size
+        ), f"set_torch_image input must be BCHW with long side {self.model.image_size}."
+        self.reset_image()
+
+        self.original_size = original_image_size
+        self.input_size = tuple(transformed_image.shape[-2:])
+        input_image = self.model.preprocess(transformed_image)
+        self.features, self.interm_feats = self.model.encoder(input_image)
+        self.is_image_set = True
+        
+    def predict(
+        self,
+        point_coords: Optional[np.ndarray] = None,
+        point_labels: Optional[np.ndarray] = None,
+        box: Optional[np.ndarray] = None,
+        multimask_output: bool = True,
+        return_logits: bool = False,
+    ) -> Tuple[np.ndarray, np.ndarray, np.ndarray]:
+        """
+        Predict masks for the given input prompts, using the currently set image.
+
+        Arguments:
+          point_coords (np.ndarray or None): A Nx2 array of point prompts to the
+            model. Each point is in (X,Y) in pixels.
+          point_labels (np.ndarray or None): A length N array of labels for the
+            point prompts. 1 indicates a foreground point and 0 indicates a
+            background point.
+          box (np.ndarray or None): A length 4 array given a box prompt to the
+            model, in XYXY format.
+          mask_input (np.ndarray): A low resolution mask input to the model, typically
+            coming from a previous prediction iteration. Has form 1xHxW, where
+            for SAM, H=W=256.
+          multimask_output (bool): If true, the model will return three masks.
+            For ambiguous input prompts (such as a single click), this will often
+            produce better masks than a single prediction. If only a single
+            mask is needed, the model's predicted quality score can be used
+            to select the best mask. For non-ambiguous prompts, such as multiple
+            input prompts, multimask_output=False can give better results.
+          return_logits (bool): If true, returns un-thresholded masks logits
+            instead of a binary mask.
+
+        Returns:
+          (np.ndarray): The output masks in CxHxW format, where C is the
+            number of masks, and (H, W) is the original image size.
+          (np.ndarray): An array of length C containing the model's
+            predictions for the quality of each mask.
+          (np.ndarray): An array of shape CxHxW, where C is the number
+            of masks and H=W=256. These low resolution logits can be passed to
+            a subsequent iteration as mask input.
+        """
+        if not self.is_image_set:
+            raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
+
+        # Transform input prompts
+        coords_torch = None
+        labels_torch = None
+        box_torch = None
+        
+        if point_coords is not None:
+            assert (
+                point_labels is not None
+            ), "point_labels must be supplied if point_coords is supplied."
+            point_coords = self.transform.apply_coords(point_coords, self.original_size)
+            coords_torch = torch.as_tensor(point_coords, dtype=torch.float, device=self.device)
+            labels_torch = torch.as_tensor(point_labels, dtype=torch.float, device=self.device)
+            coords_torch, labels_torch = coords_torch[None, :, :], labels_torch[None, :]
+        if box is not None:
+            box = self.transform.apply_boxes(box, self.original_size)
+            box_torch = torch.as_tensor(box, dtype=torch.float, device=self.device)
+
+        masks, iou_predictions, low_res_masks = self.predict_torch(
+            coords_torch,
+            labels_torch,
+            box_torch,
+            multimask_output,
+            return_logits=return_logits,
+        )
+
+        masks_np = masks[0].detach().cpu().numpy()
+        iou_predictions_np = iou_predictions[0].detach().cpu().numpy()
+        low_res_masks_np = low_res_masks[0].detach().cpu().numpy()
+        
+        return masks_np, iou_predictions_np, low_res_masks_np
+
+    @torch.no_grad()
+    def predict_torch(
+        self,
+        point_coords: Optional[torch.Tensor],
+        point_labels: Optional[torch.Tensor],
+        boxes: Optional[torch.Tensor] = None,
+        multimask_output: bool = True,
+        return_logits: bool = False,
+    ) -> Tuple[torch.Tensor, torch.Tensor, torch.Tensor]:
+        """
+        Predict masks for the given input prompts, using the currently set image.
+        Input prompts are batched torch tensors and are expected to already be
+        transformed to the input frame using ResizeLongestSide.
+
+        Arguments:
+          point_coords (torch.Tensor or None): A BxNx2 array of point prompts to the
+            model. Each point is in (X,Y) in pixels.
+          point_labels (torch.Tensor or None): A BxN array of labels for the
+            point prompts. 1 indicates a foreground point and 0 indicates a
+            background point.
+          boxes (np.ndarray or None): A Bx4 array given a box prompt to the
+            model, in XYXY format.
+          mask_input (np.ndarray): A low resolution mask input to the model, typically
+            coming from a previous prediction iteration. Has form Bx1xHxW, where
+            for SAM, H=W=256. Masks returned by a previous iteration of the
+            predict method do not need further transformation.
+          multimask_output (bool): If true, the model will return three masks.
+            For ambiguous input prompts (such as a single click), this will often
+            produce better masks than a single prediction. If only a single
+            mask is needed, the model's predicted quality score can be used
+            to select the best mask. For non-ambiguous prompts, such as multiple
+            input prompts, multimask_output=False can give better results.
+          return_logits (bool): If true, returns un-thresholded masks logits
+            instead of a binary mask.
+
+        Returns:
+          (torch.Tensor): The output masks in BxCxHxW format, where C is the
+            number of masks, and (H, W) is the original image size.
+          (torch.Tensor): An array of shape BxC containing the model's
+            predictions for the quality of each mask.
+          (torch.Tensor): An array of shape BxCxHxW, where C is the number
+            of masks and H=W=256. These low res logits can be passed to
+            a subsequent iteration as mask input.
+        """
+        if not self.is_image_set:
+            raise RuntimeError("An image must be set with .set_image(...) before mask prediction.")
+
+        if point_coords is not None:
+            points = (point_coords, point_labels)
+            attn_mask = self.model.point_attn_mask(point_coords)
+        else:
+            points = None
+            attn_mask = self.model.bbox_attn_mask(boxes)
+
+        # Embed prompts
+        masks, iou_predictions = self.model.decoder(
+            interm_feats=self.interm_feats,
+            image_embeddings=self.features,
+            points=points,
+            boxes=boxes,
+            attn_mask=attn_mask,
+        )
+        
+        # Select the correct mask or masks for output
+        if multimask_output:
+            mask_slice = slice(0, None)
+        else:
+            mask_slice = slice(0, 1)
+            
+        masks = masks[:, mask_slice, :, :]
+        iou_predictions = iou_predictions[:, mask_slice]
+            
+        low_res_masks = F.interpolate(masks, scale_factor=2, mode='bilinear', align_corners=False)
+            
+        masks = self.model.postprocess_masks(
+            masks,
+            input_size=self.input_size,
+            original_size=self.original_size,
+        )
+        
+        return masks.sigmoid(), iou_predictions, low_res_masks.sigmoid()
+
+    def get_image_embedding(self) -> torch.Tensor:
+        """
+        Returns the image embeddings for the currently set image, with
+        shape 1xCxHxW, where C is the embedding dimension and (H,W) are
+        the embedding spatial dimension of SAM (typically C=256, H=W=64).
+        """
+        if not self.is_image_set:
+            raise RuntimeError(
+                "An image must be set with .set_image(...) to generate an embedding."
+            )
+        assert self.features is not None, "Features must exist if an image has been set."
+        return self.features
+
+    @property
+    def device(self) -> torch.device:
+        return self.model.device
+
+    def reset_image(self) -> None:
+        """Resets the currently set image."""
+        self.is_image_set = False
+        self.features = None
+        self.interm_feats = None
+        self.orig_h = None
+        self.orig_w = None
+        self.input_h = None
+        self.input_w = None
+        

zim/utils/__init__.py

@@ -0,0 +1,10 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+
+# This source code is licensed under the license found in the
+# LICENSE file in the root directory of this source tree.
+
+from .argparser import get_parser
+from .print import print_once, pretty
+from .utils import AverageMeter, ResizeLongestSide
+from .amg import show_mat_anns

zim/utils/amg.py

@@ -0,0 +1,373 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+This source code is based on code from the Segment Anything Model (SAM)
+(https://github.com/facebookresearch/segment-anything).
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import numpy as np
+import torch
+import cv2
+
+import math
+from copy import deepcopy
+from itertools import product
+from typing import Any, Dict, Generator, ItemsView, List, Tuple
+
+
+class MaskData:
+    """
+    A structure for storing masks and their related data in batched format.
+    Implements basic filtering and concatenation.
+    """
+
+    def __init__(self, **kwargs) -> None:
+        for v in kwargs.values():
+            assert isinstance(
+                v, (list, np.ndarray, torch.Tensor)
+            ), "MaskData only supports list, numpy arrays, and torch tensors."
+        self._stats = dict(**kwargs)
+
+    def __setitem__(self, key: str, item: Any) -> None:
+        assert isinstance(
+            item, (list, np.ndarray, torch.Tensor)
+        ), "MaskData only supports list, numpy arrays, and torch tensors."
+        self._stats[key] = item
+
+    def __delitem__(self, key: str) -> None:
+        del self._stats[key]
+
+    def __getitem__(self, key: str) -> Any:
+        return self._stats[key]
+
+    def items(self) -> ItemsView[str, Any]:
+        return self._stats.items()
+
+    def filter(self, keep: torch.Tensor) -> None:
+        for k, v in self._stats.items():
+            if v is None:
+                self._stats[k] = None
+            elif isinstance(v, torch.Tensor):
+                self._stats[k] = v[torch.as_tensor(keep, device=v.device)]
+            elif isinstance(v, np.ndarray):
+                self._stats[k] = v[keep.detach().cpu().numpy()]
+            elif isinstance(v, list) and keep.dtype == torch.bool:
+                self._stats[k] = [a for i, a in enumerate(v) if keep[i]]
+            elif isinstance(v, list):
+                self._stats[k] = [v[i] for i in keep]
+            else:
+                raise TypeError(f"MaskData key {k} has an unsupported type {type(v)}.")
+
+    def cat(self, new_stats: "MaskData") -> None:
+        for k, v in new_stats.items():
+            if k not in self._stats or self._stats[k] is None:
+                self._stats[k] = deepcopy(v)
+            elif isinstance(v, torch.Tensor):
+                self._stats[k] = torch.cat([self._stats[k], v], dim=0)
+            elif isinstance(v, np.ndarray):
+                self._stats[k] = np.concatenate([self._stats[k], v], axis=0)
+            elif isinstance(v, list):
+                self._stats[k] = self._stats[k] + deepcopy(v)
+            else:
+                raise TypeError(f"MaskData key {k} has an unsupported type {type(v)}.")
+
+    def to_numpy(self) -> None:
+        for k, v in self._stats.items():
+            if isinstance(v, torch.Tensor):
+                self._stats[k] = v.detach().cpu().numpy()
+
+
+def is_box_near_crop_edge(
+    boxes: torch.Tensor, crop_box: List[int], orig_box: List[int], atol: float = 20.0
+) -> torch.Tensor:
+    """Filter masks at the edge of a crop, but not at the edge of the original image."""
+    crop_box_torch = torch.as_tensor(crop_box, dtype=torch.float, device=boxes.device)
+    orig_box_torch = torch.as_tensor(orig_box, dtype=torch.float, device=boxes.device)
+    boxes = uncrop_boxes_xyxy(boxes, crop_box).float()
+    near_crop_edge = torch.isclose(boxes, crop_box_torch[None, :], atol=atol, rtol=0)
+    near_image_edge = torch.isclose(boxes, orig_box_torch[None, :], atol=atol, rtol=0)
+    near_crop_edge = torch.logical_and(near_crop_edge, ~near_image_edge)
+    return torch.any(near_crop_edge, dim=1)
+
+
+def box_xyxy_to_xywh(box_xyxy: torch.Tensor) -> torch.Tensor:
+    box_xywh = deepcopy(box_xyxy)
+    box_xywh[2] = box_xywh[2] - box_xywh[0]
+    box_xywh[3] = box_xywh[3] - box_xywh[1]
+    return box_xywh
+
+
+def batch_iterator(batch_size: int, *args) -> Generator[List[Any], None, None]:
+    assert len(args) > 0 and all(
+        len(a) == len(args[0]) for a in args
+    ), "Batched iteration must have inputs of all the same size."
+    n_batches = len(args[0]) // batch_size + int(len(args[0]) % batch_size != 0)
+    for b in range(n_batches):
+        yield [arg[b * batch_size : (b + 1) * batch_size] for arg in args]
+
+
+def mask_to_rle_pytorch(tensor: torch.Tensor) -> List[Dict[str, Any]]:
+    """
+    Encodes masks to an uncompressed RLE, in the format expected by
+    pycoco tools.
+    """
+    # Put in fortran order and flatten h,w
+    b, h, w = tensor.shape
+    tensor = tensor.permute(0, 2, 1).flatten(1)
+
+    # Compute change indices
+    diff = tensor[:, 1:] ^ tensor[:, :-1]
+    change_indices = diff.nonzero()
+
+    # Encode run length
+    out = []
+    for i in range(b):
+        cur_idxs = change_indices[change_indices[:, 0] == i, 1]
+        cur_idxs = torch.cat(
+            [
+                torch.tensor([0], dtype=cur_idxs.dtype, device=cur_idxs.device),
+                cur_idxs + 1,
+                torch.tensor([h * w], dtype=cur_idxs.dtype, device=cur_idxs.device),
+            ]
+        )
+        btw_idxs = cur_idxs[1:] - cur_idxs[:-1]
+        counts = [] if tensor[i, 0] == 0 else [0]
+        counts.extend(btw_idxs.detach().cpu().tolist())
+        out.append({"size": [h, w], "counts": counts})
+    return out
+
+
+def rle_to_mask(rle: Dict[str, Any]) -> np.ndarray:
+    """Compute a binary mask from an uncompressed RLE."""
+    h, w = rle["size"]
+    mask = np.empty(h * w, dtype=bool)
+    idx = 0
+    parity = False
+    for count in rle["counts"]:
+        mask[idx : idx + count] = parity
+        idx += count
+        parity ^= True
+    mask = mask.reshape(w, h)
+    return mask.transpose()  # Put in C order
+
+
+def area_from_rle(rle: Dict[str, Any]) -> int:
+    return sum(rle["counts"][1::2])
+
+
+def calculate_stability_score(
+    masks: torch.Tensor, mask_threshold: float, threshold_offset: float
+) -> torch.Tensor:
+    """
+    Computes the stability score for a batch of masks. The stability
+    score is the IoU between the binary masks obtained by thresholding
+    the predicted mask logits at high and low values.
+    """
+    # One mask is always contained inside the other.
+    # Save memory by preventing unnecessary cast to torch.int64
+    intersections = (
+        (masks > (mask_threshold + threshold_offset))
+        .sum(-1, dtype=torch.int16)
+        .sum(-1, dtype=torch.int32)
+    )
+    unions = (
+        (masks > (mask_threshold - threshold_offset))
+        .sum(-1, dtype=torch.int16)
+        .sum(-1, dtype=torch.int32)
+    )
+    return intersections / unions
+
+
+def build_point_grid(n_per_side: int) -> np.ndarray:
+    """Generates a 2D grid of points evenly spaced in [0,1]x[0,1]."""
+    offset = 1 / (2 * n_per_side)
+    points_one_side = np.linspace(offset, 1 - offset, n_per_side)
+    points_x = np.tile(points_one_side[None, :], (n_per_side, 1))
+    points_y = np.tile(points_one_side[:, None], (1, n_per_side))
+    points = np.stack([points_x, points_y], axis=-1).reshape(-1, 2)
+    return points
+
+
+def build_all_layer_point_grids(
+    n_per_side: int, n_layers: int, scale_per_layer: int
+) -> List[np.ndarray]:
+    """Generates point grids for all crop layers."""
+    points_by_layer = []
+    for i in range(n_layers + 1):
+        n_points = int(n_per_side / (scale_per_layer**i))
+        points_by_layer.append(build_point_grid(n_points))
+    return points_by_layer
+
+
+def generate_crop_boxes(
+    im_size: Tuple[int, ...], n_layers: int, overlap_ratio: float
+) -> Tuple[List[List[int]], List[int]]:
+    """
+    Generates a list of crop boxes of different sizes. Each layer
+    has (2**i)**2 boxes for the ith layer.
+    """
+    crop_boxes, layer_idxs = [], []
+    im_h, im_w = im_size
+    short_side = min(im_h, im_w)
+
+    # Original image
+    crop_boxes.append([0, 0, im_w, im_h])
+    layer_idxs.append(0)
+
+    def crop_len(orig_len, n_crops, overlap):
+        return int(math.ceil((overlap * (n_crops - 1) + orig_len) / n_crops))
+
+    for i_layer in range(n_layers):
+        n_crops_per_side = 2 ** (i_layer + 1)
+        overlap = int(overlap_ratio * short_side * (2 / n_crops_per_side))
+
+        crop_w = crop_len(im_w, n_crops_per_side, overlap)
+        crop_h = crop_len(im_h, n_crops_per_side, overlap)
+
+        crop_box_x0 = [int((crop_w - overlap) * i) for i in range(n_crops_per_side)]
+        crop_box_y0 = [int((crop_h - overlap) * i) for i in range(n_crops_per_side)]
+
+        # Crops in XYWH format
+        for x0, y0 in product(crop_box_x0, crop_box_y0):
+            box = [x0, y0, min(x0 + crop_w, im_w), min(y0 + crop_h, im_h)]
+            crop_boxes.append(box)
+            layer_idxs.append(i_layer + 1)
+
+    return crop_boxes, layer_idxs
+
+
+def uncrop_boxes_xyxy(boxes: torch.Tensor, crop_box: List[int]) -> torch.Tensor:
+    x0, y0, _, _ = crop_box
+    offset = torch.tensor([[x0, y0, x0, y0]], device=boxes.device)
+    # Check if boxes has a channel dimension
+    if len(boxes.shape) == 3:
+        offset = offset.unsqueeze(1)
+    return boxes + offset
+
+
+def uncrop_points(points: torch.Tensor, crop_box: List[int]) -> torch.Tensor:
+    x0, y0, _, _ = crop_box
+    offset = torch.tensor([[x0, y0]], device=points.device)
+    # Check if points has a channel dimension
+    if len(points.shape) == 3:
+        offset = offset.unsqueeze(1)
+    return points + offset
+
+
+def uncrop_masks(
+    masks: torch.Tensor, crop_box: List[int], orig_h: int, orig_w: int
+) -> torch.Tensor:
+    x0, y0, x1, y1 = crop_box
+    if x0 == 0 and y0 == 0 and x1 == orig_w and y1 == orig_h:
+        return masks
+    # Coordinate transform masks
+    pad_x, pad_y = orig_w - (x1 - x0), orig_h - (y1 - y0)
+    pad = (x0, pad_x - x0, y0, pad_y - y0)
+    return torch.nn.functional.pad(masks, pad, value=0)
+
+
+def remove_small_regions(
+    mask: np.ndarray, area_thresh: float, mode: str
+) -> Tuple[np.ndarray, bool]:
+    """
+    Removes small disconnected regions and holes in a mask. Returns the
+    mask and an indicator of if the mask has been modified.
+    """
+    import cv2  # type: ignore
+
+    assert mode in ["holes", "islands"]
+    correct_holes = mode == "holes"
+    working_mask = (correct_holes ^ mask).astype(np.uint8)
+    n_labels, regions, stats, _ = cv2.connectedComponentsWithStats(working_mask, 8)
+    sizes = stats[:, -1][1:]  # Row 0 is background label
+    small_regions = [i + 1 for i, s in enumerate(sizes) if s < area_thresh]
+    if len(small_regions) == 0:
+        return mask, False
+    fill_labels = [0] + small_regions
+    if not correct_holes:
+        fill_labels = [i for i in range(n_labels) if i not in fill_labels]
+        # If every region is below threshold, keep largest
+        if len(fill_labels) == 0:
+            fill_labels = [int(np.argmax(sizes)) + 1]
+    mask = np.isin(regions, fill_labels)
+    return mask, True
+
+
+def coco_encode_rle(uncompressed_rle: Dict[str, Any]) -> Dict[str, Any]:
+    from pycocotools import mask as mask_utils  # type: ignore
+
+    h, w = uncompressed_rle["size"]
+    rle = mask_utils.frPyObjects(uncompressed_rle, h, w)
+    rle["counts"] = rle["counts"].decode("utf-8")  # Necessary to serialize with json
+    return rle
+
+
+def batched_mask_to_box(masks: torch.Tensor) -> torch.Tensor:
+    """
+    Calculates boxes in XYXY format around masks. Return [0,0,0,0] for
+    an empty mask. For input shape C1xC2x...xHxW, the output shape is C1xC2x...x4.
+    """
+    # torch.max below raises an error on empty inputs, just skip in this case
+    if torch.numel(masks) == 0:
+        return torch.zeros(*masks.shape[:-2], 4, device=masks.device)
+
+    # Normalize shape to CxHxW
+    shape = masks.shape
+    h, w = shape[-2:]
+    if len(shape) > 2:
+        masks = masks.flatten(0, -3)
+    else:
+        masks = masks.unsqueeze(0)
+
+    # Get top and bottom edges
+    in_height, _ = torch.max(masks, dim=-1)
+    in_height_coords = in_height * torch.arange(h, device=in_height.device)[None, :]
+    bottom_edges, _ = torch.max(in_height_coords, dim=-1)
+    in_height_coords = in_height_coords + h * (~in_height)
+    top_edges, _ = torch.min(in_height_coords, dim=-1)
+
+    # Get left and right edges
+    in_width, _ = torch.max(masks, dim=-2)
+    in_width_coords = in_width * torch.arange(w, device=in_width.device)[None, :]
+    right_edges, _ = torch.max(in_width_coords, dim=-1)
+    in_width_coords = in_width_coords + w * (~in_width)
+    left_edges, _ = torch.min(in_width_coords, dim=-1)
+
+    # If the mask is empty the right edge will be to the left of the left edge.
+    # Replace these boxes with [0, 0, 0, 0]
+    empty_filter = (right_edges < left_edges) | (bottom_edges < top_edges)
+    out = torch.stack([left_edges, top_edges, right_edges, bottom_edges], dim=-1)
+    out = out * (~empty_filter).unsqueeze(-1)
+
+    # Return to original shape
+    if len(shape) > 2:
+        out = out.reshape(*shape[:-2], 4)
+    else:
+        out = out[0]
+
+    return out
+
+def show_mat_anns(image, anns):
+    if len(anns) == 0:
+        return np.zeros_like(image) + 128
+    
+    sorted_anns = sorted(anns, key=(lambda x: x['area']), reverse=True)
+
+    image = image.astype(np.float32)
+    colorized_mat = np.zeros_like(image)
+    
+    for ann in sorted_anns:
+        color = (np.random.random(3) * 255).astype(np.float32)
+        if 'logit' in ann:
+            mat = ann['logit'].astype(np.float32) / 255.
+        else:
+            mat = ann['segmentation'].astype(np.float32)
+
+        color_mat = np.zeros_like(image) + color[None, None]
+        colorized_mat = color_mat * mat[:, :, None] + colorized_mat * (1. - mat[:, :, None])
+
+    colorized_mat = np.uint8(colorized_mat)
+
+    return colorized_mat

zim/utils/argparser.py

@@ -0,0 +1,96 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import os
+import argparse
+from config.config import config_
+
+def str2bool(v):
+    if isinstance(v, bool):
+        return v
+    if v.lower() in ("yes", "true", "t", "y", "1"):
+        return True
+    elif v.lower() in ("no", "false", "f", "n", "0"):
+        return False
+    else:
+        raise argparse.ArgumentTypeError("Boolean value expected.")
+
+
+def get_parser(verbose=False):
+    p = argparse.ArgumentParser("argparser", add_help=False)
+
+    p.add_argument(
+        "--data-root", type=str, default=config_.data_root, help="data root directory"
+    )
+    p.add_argument(
+        "--local_rank", type=int, default=int(os.getenv("LOCAL_RANK", "0"))
+    )
+    p.add_argument(
+        "--amp", type=str2bool, default=True
+    )
+    p.add_argument(
+        "--ddp", action="store_true"
+    )
+    p.add_argument(
+        "--random-seed", type=int, default=config_.random_seed
+    )
+
+    # network config
+    p.add_argument(
+        "--network-encoder",
+        type=str,
+        default=config_.network.encoder,
+        choices=["vit_b", "vit_l"],
+    )
+    p.add_argument(
+        "--network-decoder",
+        type=str,
+        default=config_.network.decoder,
+        choices=["zim", "sam"],
+    )
+    p.add_argument(
+        "--network-encode-kernel",
+        type=int,
+        default=config_.network.encode_kernel,
+    )
+    
+    # evaluation config
+    p.add_argument(
+        "--eval-workers", type=int, default=config_.eval.workers,
+    )
+    p.add_argument(
+        "--eval-image-size", type=int, default=config_.eval.image_size,
+    )
+    p.add_argument(
+        "--eval-prompt-type", type=str, default=config_.eval.prompt_type,
+    )
+    p.add_argument(
+        "--eval-model-list", type=str, default=config_.eval.model_list,
+    )
+    p.add_argument(
+        "--eval-zim-weights",
+        type=str,
+        default=config_.eval.zim_weights,
+    )
+    p.add_argument(
+        "--eval-sam-weights",
+        type=str,
+        default=config_.eval.sam_weights,
+    )
+    
+    # dataset config
+    p.add_argument(
+        "--dataset-valset", type=str, default=config_.dataset.valset,
+    )
+    p.add_argument(
+        "--dataset-data-type", type=str, default=config_.dataset.data_type,
+    )
+    p.add_argument(
+        "--dataset-data-list-txt", type=str, default=config_.dataset.data_list_txt,
+    )
+    
+    return p

zim/utils/print.py

@@ -0,0 +1,20 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import torch
+
+def print_once(message):
+    if not torch.distributed.is_initialized() or torch.distributed.get_rank() == 0:
+        print(message)
+
+def pretty(d, indent=0):
+    for key, value in d.items():
+        print_once("\t" * indent + str(key))
+        if isinstance(value, dict):
+            pretty(value, indent + 1)
+        else:
+            print_once("\t" * (indent + 1) + str(value))

zim/utils/utils.py

@@ -0,0 +1,148 @@
+"""
+Copyright (c) 2024-present Naver Cloud Corp.
+
+This source code is licensed under the license found in the
+LICENSE file in the root directory of this source tree.
+"""
+
+import numpy as np
+import torch
+from torch.nn import functional as F
+from torchvision.transforms.functional import resize, to_pil_image, InterpolationMode
+from copy import deepcopy
+from typing import Optional, Tuple, List
+
+class ResizeLongestSide:
+    """
+    Resizes images to the longest side 'target_length', as well as provides
+    methods for resizing coordinates and boxes. Provides methods for
+    transforming both numpy array and batched torch tensors.
+    """
+
+    def __init__(self, target_length: int) -> None:
+        self.target_length = target_length
+
+    def apply_image(self, image: np.ndarray) -> np.ndarray:
+        """
+        Expects a numpy array with shape HxWxC in uint8 format.
+        """
+        target_size = self.get_preprocess_shape(image.shape[0], image.shape[1], self.target_length)
+        return np.array(resize(to_pil_image(image), target_size))
+
+    def apply_coords(self, coords: np.ndarray, original_size: Tuple[int, ...]) -> np.ndarray:
+        """
+        Expects a numpy array of length 2 in the final dimension. Requires the
+        original image size in (H, W) format.
+        """
+        old_h, old_w = original_size
+        new_h, new_w = self.get_preprocess_shape(
+            original_size[0], original_size[1], self.target_length
+        )
+        coords = deepcopy(coords).astype(float)
+        coords[..., 0] = coords[..., 0] * (new_w / old_w)
+        coords[..., 1] = coords[..., 1] * (new_h / old_h)
+        return coords
+
+    def apply_boxes(self, boxes: np.ndarray, original_size: Tuple[int, ...]) -> np.ndarray:
+        """
+        Expects a numpy array shape Bx4. Requires the original image size
+        in (H, W) format.
+        """
+        boxes = self.apply_coords(boxes.reshape(-1, 2, 2), original_size)
+        return boxes.reshape(-1, 4)
+
+    def apply_image_torch(self, image: torch.Tensor) -> torch.Tensor:
+        """
+        Expects batched images with shape BxCxHxW and float format. This
+        transformation may not exactly match apply_image. apply_image is
+        the transformation expected by the model.
+        """
+        # Expects an image in BCHW format. May not exactly match apply_image.
+        target_size = self.get_preprocess_shape(image.shape[2], image.shape[3], self.target_length)
+        return F.interpolate(
+            image, target_size, mode="bilinear", align_corners=False, antialias=True
+        )
+
+    def apply_coords_torch(
+        self, coords: torch.Tensor, original_size: Tuple[int, ...]
+    ) -> torch.Tensor:
+        """
+        Expects a torch tensor with length 2 in the last dimension. Requires the
+        original image size in (H, W) format.
+        """
+        old_h, old_w = original_size
+        new_h, new_w = self.get_preprocess_shape(
+            original_size[0], original_size[1], self.target_length
+        )
+        coords = deepcopy(coords).to(torch.float)
+        coords[..., 0] = coords[..., 0] * (new_w / old_w)
+        coords[..., 1] = coords[..., 1] * (new_h / old_h)
+        return coords
+
+    def apply_boxes_torch(
+        self, boxes: torch.Tensor, original_size: Tuple[int, ...]
+    ) -> torch.Tensor:
+        """
+        Expects a torch tensor with shape Bx4. Requires the original image
+        size in (H, W) format.
+        """
+        boxes = self.apply_coords_torch(boxes.reshape(-1, 2, 2), original_size)
+        return boxes.reshape(-1, 4)
+    
+    def apply_mask(self, image: np.ndarray) -> np.ndarray:
+        """
+        Expects a numpy array with shape HxWxC in uint8 format.
+        """
+        target_size = self.get_preprocess_shape(image.shape[0], image.shape[1], self.target_length)
+        return np.array(resize(to_pil_image(image), target_size, interpolation=InterpolationMode.NEAREST))
+
+    @staticmethod
+    def get_preprocess_shape(oldh: int, oldw: int, long_side_length: int) -> Tuple[int, int]:
+        """
+        Compute the output size given input size and target long side length.
+        """
+        scale = long_side_length * 1.0 / max(oldh, oldw)
+        newh, neww = oldh * scale, oldw * scale
+        neww = int(neww + 0.5)
+        newh = int(newh + 0.5)
+        return (newh, neww)
+    
+    
+def remove_prefix(text, prefix):
+    if text.startswith(prefix):
+        return text[len(prefix) :]
+    return text
+
+class AverageMeter(object):
+    """Computes and stores the average and current value"""
+
+    def __init__(self, is_ddp):
+        self.is_ddp = is_ddp
+        self.reset()
+
+    def reset(self):
+        self.val = 0.0
+        self.avg = 0.0
+        self.sum = 0.0
+        self.count = 0.0
+
+    def update(self, val, n=1):
+        self.val = val
+        self.sum += val * n
+        self.count += n
+        self.avg = self.sum / (self.count + 1e-5)
+
+    def synch(self, device):
+        if self.is_ddp is False:
+            return
+
+        _sum = torch.tensor(self.sum).to(device)
+        _count = torch.tensor(self.count).to(device)
+
+        torch.distributed.reduce(_sum, dst=0)
+        torch.distributed.reduce(_count, dst=0)
+
+        if torch.distributed.get_rank() == 0:
+            self.sum = _sum.item()
+            self.count = _count.item()
+            self.avg = self.sum / (self.count + 1e-5)