256 lines
10 KiB
Python
256 lines
10 KiB
Python
|
# Modified by Lu He
|
|||
|
|
# ------------------------------------------------------------------------
|
|||
|
|
# Deformable DETR
|
|||
|
|
# Copyright (c) 2020 SenseTime. All Rights Reserved.
|
|||
|
|
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
|||
|
|
# ------------------------------------------------------------------------
|
|||
|
|
# Modified from DETR (https://github.com/facebookresearch/detr)
|
|||
|
|
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
|||
|
|
# ------------------------------------------------------------------------
|
|||
|
|
|
|||
|
|
"""
|
|||
|
|
COCO dataset which returns image_id for evaluation.
|
|||
|
|
|
|||
|
|
Mostly copy-paste from https://github.com/pytorch/vision/blob/13b35ff/references/detection/coco_utils.py
|
|||
|
|
"""
|
|||
|
|
from pathlib import Path
|
|||
|
|
|
|||
|
|
import torch
|
|||
|
|
import torch.utils.data
|
|||
|
|
from pycocotools import mask as coco_mask
|
|||
|
|
from .coco_video_parser import CocoVID
|
|||
|
|
from .torchvision_datasets import CocoDetection as TvCocoDetection
|
|||
|
|
from util.misc import get_local_rank, get_local_size
|
|||
|
|
import datasets.transforms_multi as T
|
|||
|
|
from torch.utils.data.dataset import ConcatDataset
|
|||
|
|
import random
|
|||
|
|
|
|||
|
|
class CocoDetection(TvCocoDetection):
|
|||
|
|
def __init__(self, img_folder, ann_file, transforms, return_masks, interval1, interval2, num_ref_frames= 3,
|
|||
|
|
is_train = True, filter_key_img=True, cache_mode=False, local_rank=0, local_size=1):
|
|||
|
|
super(CocoDetection, self).__init__(img_folder, ann_file,
|
|||
|
|
cache_mode=cache_mode, local_rank=local_rank, local_size=local_size)
|
|||
|
|
self._transforms = transforms
|
|||
|
|
self.prepare = ConvertCocoPolysToMask(return_masks)
|
|||
|
|
self.ann_file = ann_file
|
|||
|
|
self.frame_range = [-2, 2]
|
|||
|
|
self.num_ref_frames = num_ref_frames
|
|||
|
|
self.cocovid = CocoVID(self.ann_file)
|
|||
|
|
self.is_train = is_train
|
|||
|
|
self.filter_key_img = filter_key_img
|
|||
|
|
self.interval1 = interval1
|
|||
|
|
self.interval2 = interval2
|
|||
|
|
|
|||
|
|
def __getitem__(self, idx):
|
|||
|
|
"""
|
|||
|
|
Args:
|
|||
|
|
index (int): Index
|
|||
|
|
Returns:
|
|||
|
|
tuple: Tuple (image, target). target is the object returned by ``coco.loadAnns``.
|
|||
|
|
"""
|
|||
|
|
# idx若为675834，则img_id为675835(img_id=idx+1)
|
|||
|
|
imgs = []
|
|||
|
|
coco = self.coco
|
|||
|
|
img_id = self.ids[idx]
|
|||
|
|
ann_ids = coco.getAnnIds(imgIds=img_id)
|
|||
|
|
target = coco.loadAnns(ann_ids)
|
|||
|
|
img_info = coco.loadImgs(img_id)[0]
|
|||
|
|
path = img_info['file_name']
|
|||
|
|
video_id = img_info['video_id']
|
|||
|
|
img = self.get_image(path)
|
|||
|
|
target = {'image_id': img_id, 'annotations': target}
|
|||
|
|
img, target = self.prepare(img, target)
|
|||
|
|
imgs.append(img)
|
|||
|
|
if video_id == -1: # imgnet_det
|
|||
|
|
for i in range(self.num_ref_frames):
|
|||
|
|
imgs.append(img)
|
|||
|
|
else: # imgnet_vid
|
|||
|
|
img_ids = self.cocovid.get_img_ids_from_vid(video_id)
|
|||
|
|
ref_img_ids = []
|
|||
|
|
if self.is_train: # Train
|
|||
|
|
interval = self.num_ref_frames + 2 # *20
|
|||
|
|
left = max(img_ids[0], img_id - interval)
|
|||
|
|
right = min(img_ids[-1], img_id + interval)
|
|||
|
|
sample_range = list(range(left, right+1))
|
|||
|
|
if self.num_ref_frames >= 10:
|
|||
|
|
sample_range=img_ids
|
|||
|
|
if self.filter_key_img and img_id in sample_range:
|
|||
|
|
sample_range.remove(img_id)
|
|||
|
|
while len(sample_range) < self.num_ref_frames:
|
|||
|
|
print("sample_range", sample_range)
|
|||
|
|
sample_range.extend(sample_range)
|
|||
|
|
ref_img_ids = random.sample(sample_range, self.num_ref_frames)
|
|||
|
|
|
|||
|
|
else: # Eval
|
|||
|
|
ref_img_ids = []
|
|||
|
|
Len = len(img_ids)
|
|||
|
|
interval = max(int(Len // 16), 1)
|
|||
|
|
if self.num_ref_frames < 8:
|
|||
|
|
left_indexs = int((img_id - img_ids[0]) // interval)
|
|||
|
|
right_indexs = int((img_ids[-1] - img_id) // interval)
|
|||
|
|
if left_indexs < self.num_ref_frames:
|
|||
|
|
for i in range(self.num_ref_frames):
|
|||
|
|
ref_img_ids.append(min(img_id + (i+1)*interval, img_ids[-1]))
|
|||
|
|
else:
|
|||
|
|
for i in range(self.num_ref_frames):
|
|||
|
|
ref_img_ids.append(max(img_id - (i+1)* interval, img_ids[0]))
|
|||
|
|
sample_range = []
|
|||
|
|
if self.num_ref_frames >= 8:
|
|||
|
|
left_indexs = int((img_ids[0] - img_id) // interval)
|
|||
|
|
right_indexs = int((img_ids[-1] - img_id) // interval)
|
|||
|
|
for i in range(left_indexs, right_indexs):
|
|||
|
|
if i < 0:
|
|||
|
|
index = max(img_id + i*interval, img_ids[0])
|
|||
|
|
sample_range.append(index)
|
|||
|
|
elif i > 0:
|
|||
|
|
index = min(img_id + i * interval, img_ids[-1])
|
|||
|
|
sample_range.append(index)
|
|||
|
|
if self.filter_key_img and img_id in sample_range:
|
|||
|
|
sample_range.remove(img_id)
|
|||
|
|
while len(sample_range) < self.num_ref_frames:
|
|||
|
|
print("sample_range", sample_range)
|
|||
|
|
sample_range.extend(sample_range)
|
|||
|
|
ref_img_ids = sample_range[:self.num_ref_frames]
|
|||
|
|
|
|||
|
|
for ref_img_id in ref_img_ids:
|
|||
|
|
ref_ann_ids = coco.getAnnIds(imgIds=ref_img_id)
|
|||
|
|
ref_img_info = coco.loadImgs(ref_img_id)[0]
|
|||
|
|
ref_img_path = ref_img_info['file_name']
|
|||
|
|
ref_img = self.get_image(ref_img_path)
|
|||
|
|
imgs.append(ref_img)
|
|||
|
|
|
|||
|
|
if self._transforms is not None:
|
|||
|
|
imgs, target = self._transforms(imgs, target)
|
|||
|
|
|
|||
|
|
return torch.cat(imgs, dim=0), target
|
|||
|
|
|
|||
|
|
|
|||
|
|
def convert_coco_poly_to_mask(segmentations, height, width):
|
|||
|
|
masks = []
|
|||
|
|
for polygons in segmentations:
|
|||
|
|
rles = coco_mask.frPyObjects(polygons, height, width)
|
|||
|
|
mask = coco_mask.decode(rles)
|
|||
|
|
if len(mask.shape) < 3:
|
|||
|
|
mask = mask[..., None]
|
|||
|
|
mask = torch.as_tensor(mask, dtype=torch.uint8)
|
|||
|
|
mask = mask.any(dim=2)
|
|||
|
|
masks.append(mask)
|
|||
|
|
if masks:
|
|||
|
|
masks = torch.stack(masks, dim=0)
|
|||
|
|
else:
|
|||
|
|
masks = torch.zeros((0, height, width), dtype=torch.uint8)
|
|||
|
|
return masks
|
|||
|
|
|
|||
|
|
|
|||
|
|
class ConvertCocoPolysToMask(object):
|
|||
|
|
def __init__(self, return_masks=False):
|
|||
|
|
self.return_masks = return_masks
|
|||
|
|
|
|||
|
|
def __call__(self, image, target):
|
|||
|
|
w, h = image.size
|
|||
|
|
|
|||
|
|
image_id = target["image_id"]
|
|||
|
|
image_id = torch.tensor([image_id])
|
|||
|
|
|
|||
|
|
anno = target["annotations"]
|
|||
|
|
|
|||
|
|
anno = [obj for obj in anno if 'iscrowd' not in obj or obj['iscrowd'] == 0]
|
|||
|
|
|
|||
|
|
boxes = [obj["bbox"] for obj in anno]
|
|||
|
|
# guard against no boxes via resizing
|
|||
|
|
boxes = torch.as_tensor(boxes, dtype=torch.float32).reshape(-1, 4)
|
|||
|
|
boxes[:, 2:] += boxes[:, :2]
|
|||
|
|
boxes[:, 0::2].clamp_(min=0, max=w)
|
|||
|
|
boxes[:, 1::2].clamp_(min=0, max=h)
|
|||
|
|
|
|||
|
|
classes = [obj["category_id"] for obj in anno]
|
|||
|
|
classes = torch.tensor(classes, dtype=torch.int64)
|
|||
|
|
|
|||
|
|
if self.return_masks:
|
|||
|
|
segmentations = [obj["segmentation"] for obj in anno]
|
|||
|
|
masks = convert_coco_poly_to_mask(segmentations, h, w)
|
|||
|
|
|
|||
|
|
keypoints = None
|
|||
|
|
if anno and "keypoints" in anno[0]:
|
|||
|
|
keypoints = [obj["keypoints"] for obj in anno]
|
|||
|
|
keypoints = torch.as_tensor(keypoints, dtype=torch.float32)
|
|||
|
|
num_keypoints = keypoints.shape[0]
|
|||
|
|
if num_keypoints:
|
|||
|
|
keypoints = keypoints.view(num_keypoints, -1, 3)
|
|||
|
|
|
|||
|
|
keep = (boxes[:, 3] > boxes[:, 1]) & (boxes[:, 2] > boxes[:, 0])
|
|||
|
|
boxes = boxes[keep]
|
|||
|
|
classes = classes[keep]
|
|||
|
|
if self.return_masks:
|
|||
|
|
masks = masks[keep]
|
|||
|
|
if keypoints is not None:
|
|||
|
|
keypoints = keypoints[keep]
|
|||
|
|
|
|||
|
|
target = {}
|
|||
|
|
target["boxes"] = boxes
|
|||
|
|
target["labels"] = classes
|
|||
|
|
if self.return_masks:
|
|||
|
|
target["masks"] = masks
|
|||
|
|
target["image_id"] = image_id
|
|||
|
|
if keypoints is not None:
|
|||
|
|
target["keypoints"] = keypoints
|
|||
|
|
|
|||
|
|
# for conversion to coco api
|
|||
|
|
area = torch.tensor([obj["area"] for obj in anno])
|
|||
|
|
iscrowd = torch.tensor([obj["iscrowd"] if "iscrowd" in obj else 0 for obj in anno])
|
|||
|
|
target["area"] = area[keep]
|
|||
|
|
target["iscrowd"] = iscrowd[keep]
|
|||
|
|
|
|||
|
|
target["orig_size"] = torch.as_tensor([int(h), int(w)])
|
|||
|
|
target["size"] = torch.as_tensor([int(h), int(w)])
|
|||
|
|
|
|||
|
|
return image, target
|
|||
|
|
|
|||
|
|
|
|||
|
|
def make_coco_transforms(image_set):
|
|||
|
|
|
|||
|
|
normalize = T.Compose([
|
|||
|
|
T.ToTensor(),
|
|||
|
|
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
|
|||
|
|
])
|
|||
|
|
|
|||
|
|
scales = [480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800]
|
|||
|
|
|
|||
|
|
if image_set == 'train_vid' or image_set == "train_det" or image_set == "train_joint":
|
|||
|
|
return T.Compose([
|
|||
|
|
T.RandomHorizontalFlip(),
|
|||
|
|
T.RandomResize([600], max_size=1000),
|
|||
|
|
normalize,
|
|||
|
|
])
|
|||
|
|
|
|||
|
|
if image_set == 'val':
|
|||
|
|
return T.Compose([
|
|||
|
|
T.RandomResize([600], max_size=1000),
|
|||
|
|
normalize,
|
|||
|
|
])
|
|||
|
|
|
|||
|
|
raise ValueError(f'unknown {image_set}')
|
|||
|
|
|
|||
|
|
|
|||
|
|
def build(image_set, args):
|
|||
|
|
root = Path(args.vid_path)
|
|||
|
|
assert root.exists(), f'provided COCO path {root} does not exist'
|
|||
|
|
mode = 'instances'
|
|||
|
|
PATHS = {
|
|||
|
|
"train_det": [(root / "Data" / "DET", root / "annotations" / 'imagenet_det_30plus1cls_vid_train.json')],
|
|||
|
|
"train_vid": [(root / "Data" / "VID", root / "annotations" / 'imagenet_vid_train.json')],
|
|||
|
|
"train_joint": [(root / "Data" , root / "annotations" / 'imagenet_vid_train_joint_30.json')],
|
|||
|
|
"val": [(root / "Data" / "VID", root / "annotations" / 'imagenet_vid_val.json')],
|
|||
|
|
}
|
|||
|
|
datasets = []
|
|||
|
|
for (img_folder, ann_file) in PATHS[image_set]:
|
|||
|
|
dataset = CocoDetection(img_folder, ann_file, transforms=make_coco_transforms(image_set), is_train =(not args.eval), interval1=args.interval1,
|
|||
|
|
interval2=args.interval2, num_ref_frames = args.num_ref_frames, return_masks=args.masks, cache_mode=args.cache_mode,
|
|||
|
|
local_rank=get_local_rank(), local_size=get_local_size())
|
|||
|
|
datasets.append(dataset)
|
|||
|
|
if len(datasets) == 1:
|
|||
|
|
return datasets[0]
|
|||
|
|
return ConcatDataset(datasets)
|
|||
|
|
|
|||
|
|
|