model/vision/grit_src/image_dense_captions.py

import sys

from detectron2.config import get_cfg

sys.path.insert(
    0, 'model/vision/grit_src/third_party/CenterNet2/projects/CenterNet2/')
from model.vision.grit_src.third_party.CenterNet2.projects.CenterNet2.centernet.config import add_centernet_config
from model.vision.grit_src.grit.config import add_grit_config

from model.vision.grit_src.grit.predictor import VisualizationDemo

# constants
WINDOW_NAME = "GRiT"


def dense_pred_to_caption_no_bbox(predictions):
    object_description = predictions["instances"].pred_object_descriptions.data
    new_caption = ""
    for i in range(len(object_description) - 1):
        new_caption += (object_description[i] + ", ")
    new_caption += (object_description[-1] + ".")
    return new_caption


def dense_pred_to_caption(predictions):
    boxes = predictions["instances"].pred_boxes if predictions[
        "instances"].has("pred_boxes") else None
    object_description = predictions["instances"].pred_object_descriptions.data
    new_caption = ""
    for i in range(len(object_description)):
        new_caption += (object_description[i] + ": " + str(
            [int(a)
             for a in boxes[i].tensor.cpu().detach().numpy()[0]])) + "; "
    return new_caption


def setup_cfg(args):
    cfg = get_cfg()
    if args["cpu"]:
        cfg.MODEL.DEVICE = "cpu"
    add_centernet_config(cfg)
    add_grit_config(cfg)
    cfg.merge_from_file(args["config_file"])
    cfg.merge_from_list(args["opts"])
    # Set score_threshold for builtin models
    cfg.MODEL.ROI_HEADS.SCORE_THRESH_TEST = args["confidence_threshold"]
    cfg.MODEL.PANOPTIC_FPN.COMBINE.INSTANCES_CONFIDENCE_THRESH = args[
        "confidence_threshold"]
    if args["test_task"]:
        cfg.MODEL.TEST_TASK = args["test_task"]
    cfg.MODEL.BEAM_SIZE = 1
    cfg.MODEL.ROI_HEADS.SOFT_NMS_ENABLED = False
    cfg.USE_ACT_CHECKPOINT = False
    cfg.freeze()
    return cfg


def get_parser(device):
    arg_dict = {
        'config_file':
        "model/vision/grit_src/configs/GRiT_B_DenseCap_ObjectDet.yaml",
        'cpu':
        False,
        'confidence_threshold':
        0.5,
        'test_task':
        'DenseCap',
        'opts':
        ["MODEL.WEIGHTS", "pretrained_models/grit_b_densecap_objectdet.pth"]
    }
    if device == "cpu":
        arg_dict["cpu"] = True
    return arg_dict


def image_caption_api(cv2_img, device='cuda'):
    args2 = get_parser(device)
    cfg = setup_cfg(args2)
    demo = VisualizationDemo(cfg)

    predictions, _ = demo.run_on_image(cv2_img)
    new_caption = dense_pred_to_caption_no_bbox(predictions)

    return new_caption