# Copyright (c) OpenMMLab. All rights reserved. from typing import Optional, Tuple, Union import mmcv import numpy as np import pycocotools.mask as maskUtils import torch from mmcv.transforms import BaseTransform from mmcv.transforms import LoadAnnotations as MMCV_LoadAnnotations from mmcv.transforms import LoadImageFromFile from mmengine.fileio import get from mmengine.structures import BaseDataElement from mmdet.registry import TRANSFORMS from mmdet.structures.bbox import get_box_type from mmdet.structures.bbox.box_type import autocast_box_type from mmdet.structures.mask import BitmapMasks, PolygonMasks @TRANSFORMS.register_module() class LoadAnnotations(MMCV_LoadAnnotations): """Load and process the ``instances`` and ``seg_map`` annotation provided by dataset. The annotation format is as the following: .. code-block:: python { 'instances': [ { # List of 4 numbers representing the bounding box of the # instance, in (x1, y1, x2, y2) order. 'bbox': [x1, y1, x2, y2], # Label of image classification. 'bbox_label': 1, # Used in instance/panoptic segmentation. The segmentation mask # of the instance or the information of segments. # 1. If list[list[float]], it represents a list of polygons, # one for each connected component of the object. Each # list[float] is one simple polygon in the format of # [x1, y1, ..., xn, yn] (n≥3). The Xs and Ys are absolute # coordinates in unit of pixels. # 2. If dict, it represents the per-pixel segmentation mask in # COCO’s compressed RLE format. The dict should have keys # “size” and “counts”. Can be loaded by pycocotools 'mask': list[list[float]] or dict, } ] # Filename of semantic or panoptic segmentation ground truth file. 'seg_map_path': 'a/b/c' } After this module, the annotation has been changed to the format below: .. code-block:: python { # In (x1, y1, x2, y2) order, float type. N is the number of bboxes # in an image 'gt_bboxes': BaseBoxes(N, 4) # In int type. 'gt_bboxes_labels': np.ndarray(N, ) # In built-in class 'gt_masks': PolygonMasks (H, W) or BitmapMasks (H, W) # In uint8 type. 'gt_seg_map': np.ndarray (H, W) # in (x, y, v) order, float type. } Required Keys: - height - width - instances - bbox (optional) - bbox_label - mask (optional) - ignore_flag - seg_map_path (optional) Added Keys: - gt_bboxes (BaseBoxes[torch.float32]) - gt_bboxes_labels (np.int64) - gt_masks (BitmapMasks | PolygonMasks) - gt_seg_map (np.uint8) - gt_ignore_flags (bool) Args: with_bbox (bool): Whether to parse and load the bbox annotation. Defaults to True. with_label (bool): Whether to parse and load the label annotation. Defaults to True. with_mask (bool): Whether to parse and load the mask annotation. Default: False. with_seg (bool): Whether to parse and load the semantic segmentation annotation. Defaults to False. poly2mask (bool): Whether to convert mask to bitmap. Default: True. box_type (str): The box type used to wrap the bboxes. If ``box_type`` is None, gt_bboxes will keep being np.ndarray. Defaults to 'hbox'. imdecode_backend (str): The image decoding backend type. The backend argument for :func:``mmcv.imfrombytes``. See :fun:``mmcv.imfrombytes`` for details. Defaults to 'cv2'. backend_args (dict, optional): Arguments to instantiate the corresponding backend. Defaults to None. """ def __init__(self, with_mask: bool = False, poly2mask: bool = True, box_type: str = 'hbox', **kwargs) -> None: super(LoadAnnotations, self).__init__(**kwargs) self.with_mask = with_mask self.poly2mask = poly2mask self.box_type = box_type def _load_bboxes(self, results: dict) -> None: """Private function to load bounding box annotations. Args: results (dict): Result dict from :obj:``mmengine.BaseDataset``. Returns: dict: The dict contains loaded bounding box annotations. """ gt_bboxes = [] gt_ignore_flags = [] for instance in results.get('instances', []): gt_bboxes.append(instance['bbox']) gt_ignore_flags.append(instance['ignore_flag']) if self.box_type is None: results['gt_bboxes'] = np.array( gt_bboxes, dtype=np.float32).reshape((-1, 4)) else: _, box_type_cls = get_box_type(self.box_type) results['gt_bboxes'] = box_type_cls(gt_bboxes, dtype=torch.float32) results['gt_ignore_flags'] = np.array(gt_ignore_flags, dtype=bool) def _load_labels(self, results: dict) -> None: """Private function to load label annotations. Args: results (dict): Result dict from :obj:``mmengine.BaseDataset``. Returns: dict: The dict contains loaded label annotations. """ gt_bboxes_labels = [] for instance in results.get('instances', []): gt_bboxes_labels.append(instance['bbox_label']) # TODO: Inconsistent with mmcv, consider how to deal with it later. results['gt_bboxes_labels'] = np.array( gt_bboxes_labels, dtype=np.int64) def _poly2mask(self, mask_ann: Union[list, dict], img_h: int, img_w: int) -> np.ndarray: """Private function to convert masks represented with polygon to bitmaps. Args: mask_ann (list | dict): Polygon mask annotation input. img_h (int): The height of output mask. img_w (int): The width of output mask. Returns: np.ndarray: The decode bitmap mask of shape (img_h, img_w). """ if isinstance(mask_ann, list): # polygon -- a single object might consist of multiple parts # we merge all parts into one mask rle code rles = maskUtils.frPyObjects(mask_ann, img_h, img_w) rle = maskUtils.merge(rles) elif isinstance(mask_ann['counts'], list): # uncompressed RLE rle = maskUtils.frPyObjects(mask_ann, img_h, img_w) else: # rle rle = mask_ann mask = maskUtils.decode(rle) return mask def _process_masks(self, results: dict) -> list: """Process gt_masks and filter invalid polygons. Args: results (dict): Result dict from :obj:``mmengine.BaseDataset``. Returns: list: Processed gt_masks. """ gt_masks = [] gt_ignore_flags = [] for instance in results.get('instances', []): gt_mask = instance['mask'] # If the annotation of segmentation mask is invalid, # ignore the whole instance. if isinstance(gt_mask, list): gt_mask = [ np.array(polygon) for polygon in gt_mask if len(polygon) % 2 == 0 and len(polygon) >= 6 ] if len(gt_mask) == 0: # ignore this instance and set gt_mask to a fake mask instance['ignore_flag'] = 1 gt_mask = [np.zeros(6)] elif not self.poly2mask: # `PolygonMasks` requires a ploygon of format List[np.array], # other formats are invalid. instance['ignore_flag'] = 1 gt_mask = [np.zeros(6)] elif isinstance(gt_mask, dict) and \ not (gt_mask.get('counts') is not None and gt_mask.get('size') is not None and isinstance(gt_mask['counts'], (list, str))): # if gt_mask is a dict, it should include `counts` and `size`, # so that `BitmapMasks` can uncompressed RLE instance['ignore_flag'] = 1 gt_mask = [np.zeros(6)] gt_masks.append(gt_mask) # re-process gt_ignore_flags gt_ignore_flags.append(instance['ignore_flag']) results['gt_ignore_flags'] = np.array(gt_ignore_flags, dtype=bool) return gt_masks def _load_masks(self, results: dict) -> None: """Private function to load mask annotations. Args: results (dict): Result dict from :obj:``mmengine.BaseDataset``. """ h, w = results['ori_shape'] gt_masks = self._process_masks(results) if self.poly2mask: gt_masks = BitmapMasks( [self._poly2mask(mask, h, w) for mask in gt_masks], h, w) else: # fake polygon masks will be ignored in `PackDetInputs` gt_masks = PolygonMasks([mask for mask in gt_masks], h, w) results['gt_masks'] = gt_masks def transform(self, results: dict) -> dict: """Function to load multiple types annotations. Args: results (dict): Result dict from :obj:``mmengine.BaseDataset``. Returns: dict: The dict contains loaded bounding box, label and semantic segmentation. """ if self.with_bbox: self._load_bboxes(results) if self.with_label: self._load_labels(results) if self.with_mask: self._load_masks(results) if self.with_seg: self._load_seg_map(results) return results def __repr__(self) -> str: repr_str = self.__class__.__name__ repr_str += f'(with_bbox={self.with_bbox}, ' repr_str += f'with_label={self.with_label}, ' repr_str += f'with_mask={self.with_mask}, ' repr_str += f'with_seg={self.with_seg}, ' repr_str += f'poly2mask={self.poly2mask}, ' repr_str += f"imdecode_backend='{self.imdecode_backend}', " repr_str += f'backend_args={self.backend_args})' return repr_str