import logging import math from typing import List import numpy as np import torch import torch.distributed as dist import torch.nn.functional as F from torch import nn from detectron2.layers import ShapeSpec from detectron2.modeling import META_ARCH_REGISTRY, build_backbone, detector_postprocess from detectron2.modeling.roi_heads import build_roi_heads from detectron2.structures import Boxes, ImageList, Instances from detectron2.utils.logger import log_first_n from fvcore.nn import giou_loss, smooth_l1_loss from .loss import SetCriterion, HungarianMatcher from .head import DynamicHead from .util.box_ops import box_cxcywh_to_xyxy, box_xyxy_to_cxcywh from .util.misc import (NestedTensor, nested_tensor_from_tensor_list, accuracy, get_world_size, interpolate, is_dist_avail_and_initialized) from detectron2.layers import Conv2d, get_norm from .MaskEncoding import PCAMaskEncoding from detectron2.modeling.backbone import PatchEmbed __all__ = ["SWINTS"] class ImgFeatExtractor(nn.Module): def __init__(self, cfg): super().__init__() # self.img_feat_layer = nn.AdaptiveAvgPool2d(1) self.cfg = cfg def forward(self, features): for i, f in enumerate(features): if i == 0: x = torch.mean(torch.mean(f, -1), -1) #self.img_feat_layer(f) else: x_p = torch.mean(torch.mean(f, -1), -1) #self.img_feat_layer(f) x = x + x_p img_feats = x.squeeze(-1).squeeze(-1).unsqueeze(1).repeat(1, self.cfg.MODEL.SWINTS.NUM_PROPOSALS, 1,) del x_p del x return img_feats @META_ARCH_REGISTRY.register() class SWINTS(nn.Module): def __init__(self, cfg): super().__init__() self.cfg = cfg self.device = torch.device(cfg.MODEL.DEVICE) self.in_features = cfg.MODEL.ROI_HEADS.IN_FEATURES self.num_classes = cfg.MODEL.SWINTS.NUM_CLASSES self.num_proposals = cfg.MODEL.SWINTS.NUM_PROPOSALS self.hidden_dim = cfg.MODEL.SWINTS.HIDDEN_DIM self.num_heads = cfg.MODEL.SWINTS.NUM_HEADS # Build Backbone. self.backbone = build_backbone(cfg) self.size_divisibility = self.backbone.size_divisibility # Build Proposals. self.pos_embeddings = nn.Embedding(self.num_proposals, self.hidden_dim) self.init_proposal_boxes = nn.Embedding(self.num_proposals, 4) nn.init.constant_(self.init_proposal_boxes.weight[:, :2], 0.5) nn.init.constant_(self.init_proposal_boxes.weight[:, 2:], 1.0) # -------- self.IFE = ImgFeatExtractor(cfg) self.mask_encoding = PCAMaskEncoding(cfg) # encoding parameters. components_path = cfg.MODEL.SWINTS.PATH_COMPONENTS # update parameters. parameters = np.load(components_path) components = nn.Parameter(torch.from_numpy(parameters['components_c'][0]).float().to(self.device),requires_grad=False) explained_variances = nn.Parameter(torch.from_numpy(parameters['explained_variance_c'][0]).float().to(self.device), requires_grad=False) means = nn.Parameter(torch.from_numpy(parameters['mean_c'][0]).float().to(self.device),requires_grad=False) self.mask_encoding.components = components self.mask_encoding.explained_variances = explained_variances self.mask_encoding.means = means # Build Dynamic Head. self.head = DynamicHead(cfg=cfg, roi_input_shape=self.backbone.output_shape()) # Loss parameters: class_weight = cfg.MODEL.SWINTS.CLASS_WEIGHT giou_weight = cfg.MODEL.SWINTS.GIOU_WEIGHT l1_weight = cfg.MODEL.SWINTS.L1_WEIGHT rec_weight = cfg.MODEL.SWINTS.REC_WEIGHT no_object_weight = cfg.MODEL.SWINTS.NO_OBJECT_WEIGHT mask_weight = cfg.MODEL.SWINTS.MASK_WEIGHT self.deep_supervision = cfg.MODEL.SWINTS.DEEP_SUPERVISION # Build Criterion. matcher = HungarianMatcher(cfg=cfg, cost_class=class_weight, cost_bbox=l1_weight, cost_giou=giou_weight, cost_mask=mask_weight) self.matcher = matcher weight_dict = {"loss_ce": class_weight, "loss_bbox": l1_weight, "loss_giou": giou_weight, "loss_feat": mask_weight, "loss_dice": mask_weight} if self.deep_supervision: aux_weight_dict = {} for i in range(self.num_heads - 1): aux_weight_dict.update({k + f"_{i}": v for k, v in weight_dict.items()}) weight_dict.update(aux_weight_dict) weight_dict["loss_rec"] = rec_weight losses = ["labels", "boxes", "masks", "rec"] self.criterion = SetCriterion(cfg=cfg, num_classes=self.num_classes, matcher=matcher, weight_dict=weight_dict, eos_coef=no_object_weight, losses=losses) pixel_mean = torch.Tensor(cfg.MODEL.PIXEL_MEAN).to(self.device).view(3, 1, 1) pixel_std = torch.Tensor(cfg.MODEL.PIXEL_STD).to(self.device).view(3, 1, 1) self.normalizer = lambda x: (x - pixel_mean) / pixel_std self.to(self.device) def forward(self, batched_inputs): """ Args: batched_inputs: a list, batched outputs of :class:`DatasetMapper` . Each item in the list contains the inputs for one image. For now, each item in the list is a dict that contains: * image: Tensor, image in (C, H, W) format. * instances: Instances Other information that's included in the original dicts, such as: * "height", "width" (int): the output resolution of the model, used in inference. See :meth:`postprocess` for details. """ images, images_whwh = self.preprocess_image(batched_inputs) if isinstance(images, (list, torch.Tensor)): images = nested_tensor_from_tensor_list(images) # Feature Extraction. src = self.backbone(images.tensor) features = list() for f in self.in_features: feature = src[f] features.append(feature) # Prepare Proposals. proposal_boxes = self.init_proposal_boxes.weight.clone() proposal_boxes = box_cxcywh_to_xyxy(proposal_boxes) proposal_boxes = proposal_boxes[None] * images_whwh[:, None, :] img_feats = self.IFE(features) bs = len(features[0]) pos_embeddings = self.pos_embeddings.weight[None].repeat(bs, 1, 1) proposal_feats = img_feats + pos_embeddings del img_feats if self.training: gt_instances = [x["instances"].to(self.device) for x in batched_inputs] targets = self.prepare_targets(gt_instances) outputs_class, outputs_coord, outputs_mask,out_rec = self.head(features, proposal_boxes, proposal_feats, targets, mask_encoding=self.mask_encoding, matcher=self.matcher) output = {'pred_logits': outputs_class[-1], 'pred_boxes': outputs_coord[-1], 'pred_masks': outputs_mask[-1], 'pred_rec': out_rec} if self.deep_supervision: output['aux_outputs'] = [{'pred_logits': a, 'pred_boxes': b, 'pred_masks': c} for a, b, c in zip(outputs_class[:-1], outputs_coord[:-1], outputs_mask[:-1])] loss_dict = self.criterion(output, targets, self.mask_encoding) weight_dict = self.criterion.weight_dict for k in loss_dict.keys(): if k in weight_dict: loss_dict[k] *= weight_dict[k] return loss_dict else: outputs_class, outputs_coord, outputs_mask,out_rec = self.head(features, proposal_boxes, proposal_feats, mask_encoding=self.mask_encoding) output = {'pred_logits': outputs_class[-1], 'pred_boxes': outputs_coord[-1], 'pred_masks': outputs_mask[-1]} box_cls = output["pred_logits"] box_pred = output["pred_boxes"] mask_pred = output["pred_masks"].unsqueeze(dim=2) results = Instances(images.image_sizes[0]) results.pred_boxes = Boxes(box_pred) results.scores = box_cls results.pred_masks = mask_pred.squeeze(1) results.pred_rec = out_rec results = [results] processed_results = [] for results_per_image, input_per_image, image_size in zip(results, batched_inputs, images.image_sizes): height = input_per_image.get("height", image_size[0]) width = input_per_image.get("width", image_size[1]) r = detector_postprocess(results_per_image, height, width) processed_results.append({"instances": r}) return processed_results @torch.no_grad() def prepare_targets(self, targets): new_targets = [] for targets_per_image in targets: target = {} h, w = targets_per_image.image_size image_size_xyxy = torch.as_tensor([w, h, w, h], dtype=torch.float, device=self.device) gt_classes = targets_per_image.gt_classes gt_boxes = targets_per_image.gt_boxes.tensor / image_size_xyxy gt_boxes = box_xyxy_to_cxcywh(gt_boxes) target["labels"] = gt_classes.to(self.device) target["boxes"] = gt_boxes.to(self.device) target["boxes_xyxy"] = targets_per_image.gt_boxes.tensor.to(self.device) target["image_size_xyxy"] = image_size_xyxy.to(self.device) image_size_xyxy_tgt = image_size_xyxy.unsqueeze(0).repeat(len(gt_boxes), 1) target["image_size_xyxy_tgt"] = image_size_xyxy_tgt.to(self.device) target["area"] = targets_per_image.gt_boxes.area().to(self.device) target["gt_masks"] = targets_per_image.gt_masks.to(self.device) masks = target['gt_masks'].crop_and_resize(targets_per_image.gt_boxes, 28) target["gt_masks"] = masks.float() target["rec"] = targets_per_image.rec.to(self.device) new_targets.append(target) return new_targets @torch.no_grad() def inference(self, box_cls, box_pred, mask_pred, image_sizes, recred): """ Arguments: box_cls (Tensor): tensor of shape (batch_size, num_proposals, K). The tensor predicts the classification probability for each proposal. box_pred (Tensor): tensors of shape (batch_size, num_proposals, 4). The tensor predicts 4-vector (x,y,w,h) box regression values for every proposal image_sizes (List[torch.Size]): the input image sizes Returns: results (List[Instances]): a list of #images elements. """ assert len(box_cls) == len(image_sizes) results = [] # scores = torch.sigmoid(box_cls) labels = torch.arange(self.num_classes, device=self.device).\ unsqueeze(0).repeat(self.num_proposals, 1).flatten(0, 1) for i, (scores_per_image, box_pred_per_image, mask_pred_per_image, image_size, rec_per_image) in enumerate(zip( scores, box_pred, mask_pred, image_sizes, rec_pred )): result = Instances(image_size) scores_per_image, topk_indices = scores_per_image.flatten(0, 1).topk(self.num_proposals, sorted=False) labels_per_image = labels[topk_indices] result.pred_boxes = Boxes(box_pred_per_image) result.scores = scores_per_image result.pred_classes = labels_per_image result.pred_masks = mask_pred_per_image result.pred_rec = rec_per_image results.append(result) return results def preprocess_image(self, batched_inputs): """ Normalize, pad and batch the input images. """ images = [self.normalizer(x["image"].to(self.device)) for x in batched_inputs] images = ImageList.from_tensors(images, self.size_divisibility) images_whwh = list() for bi in batched_inputs: h, w = bi["image"].shape[-2:] images_whwh.append(torch.tensor([w, h, w, h], dtype=torch.float32, device=self.device)) images_whwh = torch.stack(images_whwh) return images, images_whwh