def build_GT(self):
for file in glob.glob(os.path.join(self.dataset_root, 'train', 'labels', "*.xml")):
_, boxGT, labelGT, difficult = PascalVocXmlParser(file, self.cateNames).parse(filterdiff=False)
for box, label, difficult in zip(boxGT, labelGT, difficult):
self.rec_gt[file].append({
'label': label,
'bbox': box,
'difficult': difficult
})
def build_GT(self):
filepath = os.path.join(self.dataset_root, 'VOC2007', 'ImageSets', 'Main', 'test.txt')
with open(filepath, 'r') as f:
filelist = f.readlines()
filelist = [file.strip() for file in filelist]
for file in filelist:
_, boxGT, labelGT, difficult = PascalVocXmlParser(self._annopath.format(file), self.cateNames).parse(filterdiff=False)
for box, label, difficult in zip(boxGT, labelGT, difficult):
self.rec_gt[file].append({
'label': label,
'bbox': box,
'difficult': difficult
})
def _parse_annotation(self,itemidx,random_trainsize):
rootpath, filename = self._ids[itemidx]
annpath = self._annopath.format(rootpath, filename)
imgpath = self._imgpath.format(rootpath, filename)
fname, bboxes, labels = PascalVocXmlParser(annpath, self.labels).parse()
img = cv2.imread(imgpath, cv2.IMREAD_COLOR)
if self.istrain:
img, bboxes = dataAug.random_horizontal_flip(np.copy(img), np.copy(bboxes))
img, bboxes = dataAug.random_crop(np.copy(img), np.copy(bboxes))
img, bboxes = dataAug.random_translate(np.copy(img), np.copy(bboxes))
ori_shape=img.shape[:2]
img, bboxes = dataAug.img_preprocess2(np.copy(img), np.copy(bboxes),
(random_trainsize, random_trainsize), True)
return img,bboxes,labels,imgpath,ori_shape
def append(self, imgpath,annpath,nms_boxes,nms_scores,nms_labels,visualize=True):
imgpath = imgpath.decode('UTF-8')
annpath = annpath.decode('UTF-8')
if nms_boxes is not None: # do have bboxes
for i in range(nms_boxes.shape[0]):
rec = {
"img_idx": imgpath.split('/')[-1].split('.')[0],
"bbox": nms_boxes[i],
"score": float(nms_scores[i])
}
self.rec_pred[nms_labels[i]].append(rec)
if visualize and len(self.visual_imgs) < self.num_visual:
_, boxGT, labelGT, _ = PascalVocXmlParser(str(annpath), self.cateNames).parse()
boxGT=np.array(boxGT)
labelGT=np.array(labelGT)
self.append_visulize(imgpath, nms_boxes, nms_labels, nms_scores, boxGT, labelGT)
def _load_batch(self, idx_batch, random_trainsize):
img_batch = []
imgpath_batch = []
annpath_batch = []
pad_scale_batch = []
ori_shape_batch = []
grid0_batch = []
grid1_batch = []
grid2_batch = []
for idx in range(self.batch_size):
rootpath, filename = self._ids[idx_batch * self.batch_size + idx]
annpath = self._annopath.format(rootpath, filename)
imgpath = self._imgpath.format(rootpath, filename)
fname, bboxes, labels, _ = PascalVocXmlParser(
annpath, self.labels).parse()
img = cv2.imread(imgpath, cv2.IMREAD_COLOR)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
ori_shape = img.shape[:2]
# Load the annotation.
img, bboxes = self._transform(random_trainsize, random_trainsize,
img, bboxes)
list_grids = transform.preprocess(bboxes,
labels,
img.shape[:2],
class_num=len(self.labels),
anchors=self.anchors)
pad_scale = (1, 1)
img_batch.append(img)
imgpath_batch.append(imgpath)
annpath_batch.append(annpath)
ori_shape_batch.append(ori_shape)
pad_scale_batch.append(pad_scale)
grid0_batch.append(list_grids[0])
grid1_batch.append(list_grids[1])
grid2_batch.append(list_grids[2])
return np.array(img_batch).astype(np.float32), \
imgpath_batch, \
annpath_batch, \
np.array(pad_scale_batch).astype(np.float32), \
np.array(ori_shape_batch).astype(np.float32), \
np.array(grid0_batch).astype(np.float32), \
np.array(grid1_batch).astype(np.float32), \
np.array(grid2_batch).astype(np.float32)
def __getitem__(self, idx):
fname, bboxes, labels = PascalVocXmlParser(self.anns[idx],
self.labels).parse()
img = cv2.imread(fname, cv2.IMREAD_COLOR)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
ori_shape = img.shape[:2]
# Load the annotation.
img, bboxes = self._transform(img, bboxes)
list_grids = transform.preprocess(bboxes,
labels,
img.shape[:2],
class_num=1,
anchors=self.anchors)
pad_scale = (1, 1)
return img.astype(np.float32), \
fname, \
np.array(pad_scale).astype(np.float32), \
np.array(ori_shape).astype(np.float32), \
list_grids[0].astype(np.float32), \
list_grids[1].astype(np.float32), \
list_grids[2].astype(np.float32),
"car",
"cat",
"chair",
"cow",
"diningtable",
"dog",
"horse",
"motorbike",
"person",
"pottedplant",
"sheep",
"sofa",
"train",
"tvmonitor"
]
filelist = [file.strip() for file in filelist]
rec_gt = defaultdict(list)
for file in filelist:
_, boxGT, labelGT, difficult = PascalVocXmlParser(_annopath.format(file), cateNames).parse()
for box, label, difficult in zip(boxGT, labelGT, difficult):
rec_gt[file].append({
'label': label,
'bbox': box,
'difficult': difficult
})
cls="person"
img_idxs = ['000001']
for imgidx in set(img_idxs):
print(rec_gt[imgidx])
_rec = [rec for rec in rec_gt[imgidx] if rec['label'] == cls]
print(_rec)
def _load_batch(self, idx_batch, random_trainsize):
outputshapes = random_trainsize // self.strides
batch_image = np.zeros((self.batch_size, random_trainsize, random_trainsize, 3))
batch_label_sbbox = np.zeros((self.batch_size, outputshapes[0], outputshapes[0],
self._gt_per_grid, 6 + self.numcls))
batch_label_mbbox = np.zeros((self.batch_size, outputshapes[1], outputshapes[1],
self._gt_per_grid, 6 + self.numcls))
batch_label_lbbox = np.zeros((self.batch_size, outputshapes[2], outputshapes[2],
self._gt_per_grid, 6 + self.numcls))
temp_batch_sbboxes = []
temp_batch_mbboxes = []
temp_batch_lbboxes = []
imgpath_batch = []
orishape_batch = []
max_sbbox_per_img = 0
max_mbbox_per_img = 0
max_lbbox_per_img = 0
for idx in range(self.batch_size):
rootpath, filename = self._ids[idx_batch * self.batch_size + idx]
annpath = self._annopath.format(rootpath, filename)
imgpath = self._imgpath.format(rootpath, filename)
fname, bboxes, labels = PascalVocXmlParser(annpath, self.labels).parse()
img = cv2.imread(imgpath, cv2.IMREAD_COLOR)
ori_shape = img.shape[:2]
# Load the annotation.
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
img, bboxes = self._transform(random_trainsize, random_trainsize, img, bboxes)
# data augmentation in original-strongeryolo
# if self.istrain:
# img, bboxes = dataAug.random_horizontal_flip(np.copy(img), np.copy(bboxes))
# img, bboxes = dataAug.random_crop(np.copy(img), np.copy(bboxes))
# img, bboxes = dataAug.random_translate(np.copy(img), np.copy(bboxes))
# img, bboxes = dataAug.img_preprocess2(np.copy(img), np.copy(bboxes),
# (random_trainsize, random_trainsize), True)
label_sbbox, label_mbbox, label_lbbox, sbboxes, mbboxes, lbboxes = \
self.preprocess_anchorfree(bboxes, labels, outputshapes)
batch_image[idx, :, :, :] = img
batch_label_sbbox[idx, :, :, :, :] = label_sbbox
batch_label_mbbox[idx, :, :, :, :] = label_mbbox
batch_label_lbbox[idx, :, :, :, :] = label_lbbox
zeros = np.zeros((1, 4), dtype=np.float32)
sbboxes = sbboxes if len(sbboxes) != 0 else zeros
mbboxes = mbboxes if len(mbboxes) != 0 else zeros
lbboxes = lbboxes if len(lbboxes) != 0 else zeros
temp_batch_sbboxes.append(sbboxes)
temp_batch_mbboxes.append(mbboxes)
temp_batch_lbboxes.append(lbboxes)
max_sbbox_per_img = max(max_sbbox_per_img, len(sbboxes))
max_mbbox_per_img = max(max_mbbox_per_img, len(mbboxes))
max_lbbox_per_img = max(max_lbbox_per_img, len(lbboxes))
imgpath_batch.append(imgpath)
orishape_batch.append(ori_shape)
batch_sbboxes = np.array(
[np.concatenate([sbboxes, np.zeros((max_sbbox_per_img + 1 - len(sbboxes), 4), dtype=np.float32)], axis=0)
for sbboxes in temp_batch_sbboxes])
batch_mbboxes = np.array(
[np.concatenate([mbboxes, np.zeros((max_mbbox_per_img + 1 - len(mbboxes), 4), dtype=np.float32)], axis=0)
for mbboxes in temp_batch_mbboxes])
batch_lbboxes = np.array(
[np.concatenate([lbboxes, np.zeros((max_lbbox_per_img + 1 - len(lbboxes), 4), dtype=np.float32)], axis=0)
for lbboxes in temp_batch_lbboxes])
return torch.from_numpy(np.array(batch_image).transpose((0, 3, 1, 2)).astype(np.float32)), \
imgpath_batch, \
torch.from_numpy(np.array(orishape_batch).astype(np.float32)), \
torch.from_numpy(np.array(batch_label_sbbox).astype(np.float32)), \
torch.from_numpy(np.array(batch_label_mbbox).astype(np.float32)), \
torch.from_numpy(np.array(batch_label_lbbox).astype(np.float32)), \
torch.from_numpy(np.array(batch_sbboxes).astype(np.float32)), \
torch.from_numpy(np.array(batch_mbboxes).astype(np.float32)), \
torch.from_numpy(np.array(batch_lbboxes).astype(np.float32))