def build_siam_augmentation(cfg, is_train=True, modality='video'):
motion_limit = 0.0
motion_blur_prob = 0.0
compression_limit = 0.0
if is_train:
min_size = cfg.INPUT.MIN_SIZE_TRAIN
max_size = cfg.INPUT.MAX_SIZE_TRAIN
flip_horizontal_prob = 0.5 # cfg.INPUT.FLIP_PROB_TRAIN
brightness = cfg.INPUT.BRIGHTNESS
contrast = cfg.INPUT.CONTRAST
saturation = cfg.INPUT.SATURATION
hue = cfg.INPUT.HUE
if modality == 'image':
motion_limit = cfg.INPUT.MOTION_LIMIT
motion_blur_prob = cfg.INPUT.MOTION_BLUR_PROB
compression_limit = cfg.INPUT.COMPRESSION_LIMIT
else:
min_size = cfg.INPUT.MIN_SIZE_TEST
max_size = cfg.INPUT.MAX_SIZE_TEST
flip_horizontal_prob = 0.0
brightness = 0.0
contrast = 0.0
saturation = 0.0
hue = 0.0
amodal = cfg.INPUT.AMODAL
SIZE_DIVISIBILITY = cfg.DATALOADER.SIZE_DIVISIBILITY
to_bgr255 = cfg.INPUT.TO_BGR255
video_color_jitter = SiamVideoColorJitter(
brightness=brightness,
contrast=contrast,
saturation=saturation,
hue=hue,
)
normalize_transform = T.Normalize(mean=cfg.INPUT.PIXEL_MEAN,
std=cfg.INPUT.PIXEL_STD,
to_bgr255=to_bgr255)
transform = Compose([
video_color_jitter,
SiamVideoMotionBlurAugment(motion_blur_prob),
SiamVideoCompressionAugment(compression_limit),
SiamVideoMotionAugment(motion_limit, amodal),
SiamVideoResize(min_size, max_size, SIZE_DIVISIBILITY),
SiamVideoRandomHorizontalFlip(prob=flip_horizontal_prob),
# PIL image
VideoTransformer(ToTensor()),
# Torch tensor, CHW (RGB format), and range from [0, 1]
# VideoTransformer(ToBGR255(to_bgr255=to_bgr255))
VideoTransformer(normalize_transform),
])
return transform
def im_detect_bbox(model, images, target_scale, target_max_size, device):
"""
Performs bbox detection on the original image.
"""
transform = TT.Compose([
T.Resize(target_scale, target_max_size),
TT.ToTensor(),
T.Normalize(mean=cfg.INPUT.PIXEL_MEAN,
std=cfg.INPUT.PIXEL_STD,
to_bgr255=cfg.INPUT.TO_BGR255)
])
images = [transform(image) for image in images]
images = to_image_list(images, cfg.DATALOADER.SIZE_DIVISIBILITY)
return model(images.to(device))
def im_detect_bbox_hflip(model, images, target_scale, target_max_size, device):
"""
Performs bbox detection on the horizontally flipped image.
Function signature is the same as for im_detect_bbox.
"""
transform = TT.Compose([
T.Resize(target_scale, target_max_size),
TT.RandomHorizontalFlip(1.0),
TT.ToTensor(),
T.Normalize(mean=cfg.INPUT.PIXEL_MEAN,
std=cfg.INPUT.PIXEL_STD,
to_bgr255=cfg.INPUT.TO_BGR255)
])
images = [transform(image) for image in images]
images = to_image_list(images, cfg.DATALOADER.SIZE_DIVISIBILITY)
boxlists = model(images.to(device))
# Invert the detections computed on the flipped image
boxlists_inv = [boxlist.transpose(0) for boxlist in boxlists]
return boxlists_inv