def im_detect_bbox_hflip(model,
im,
target_scale,
target_max_size,
box_proposals=None):
"""Performs bbox detection on the horizontally flipped image.
Function signature is the same as for im_detect_bbox.
"""
# Compute predictions on the flipped image
im_hf = im[:, ::-1, :]
im_width = im.shape[1]
box_proposals_hf = box_utils.flip_boxes(box_proposals, im_width)
scores_hf, boxes_hf, im_scale = im_detect_bbox(model,
im_hf,
target_scale,
target_max_size,
boxes=box_proposals_hf)
if boxes_hf is not None:
# Invert the detections computed on the flipped image
boxes_inv = box_utils.flip_boxes(boxes_hf, im_width)
else:
boxes_inv = None
return scores_hf, boxes_inv, im_scale
def im_detect_bbox_hflip(model,
im,
target_scale,
target_max_size,
box_proposals=None,
region_box=None):
"""Performs bbox detection on the horizontally flipped image.
Function signature is the same as for im_detect_bbox.
"""
# Compute predictions on the flipped image
im_hf = im[:, ::-1, :]
im_width = im.shape[1]
if not cfg.MODEL.FASTER_RCNN:
box_proposals_hf = box_utils.flip_boxes(box_proposals, im_width)
else:
box_proposals_hf = None
scores_hf, boxes_hf, im_scale, _ = im_detect_bbox(model,
im_hf,
target_scale,
target_max_size,
boxes=box_proposals_hf,
region_box=region_box)
# Invert the detections computed on the flipped image
boxes_inv = box_utils.flip_boxes(boxes_hf, im_width)
return scores_hf, boxes_inv, im_scale
def __call__(self, to_transform):
if to_transform['hflip']:
to_transform['img'] = to_transform['img'][:, ::-1, :]
im_width = to_transform['img'].shape[1]
to_transform['proposals'] = box_utils.flip_boxes(
to_transform['proposals'], im_width)
return to_transform
def im_detect_bbox_hflip(model, im, box_proposals=None):
"""Performs bbox detection on the horizontally flipped image.
Function signature is the same as for im_detect_bbox.
"""
# Compute predictions on the flipped image
im_hf = im[:, ::-1, :]
im_width = im.shape[1]
if not cfg.MODEL.FASTER_RCNN:
box_proposals_hf = box_utils.flip_boxes(box_proposals, im_width)
else:
box_proposals_hf = None
scores_hf, boxes_hf, im_scales = im_detect_bbox(
model, im_hf, box_proposals_hf
)
# Invert the detections computed on the flipped image
boxes_inv = box_utils.flip_boxes(boxes_hf, im_width)
return scores_hf, boxes_inv, im_scales
def im_detect_bbox_hflip(model, im, box_proposals=None):
"""Performs bbox detection on the horizontally flipped image.
Function signature is the same as for im_detect_bbox.
"""
# Compute predictions on the flipped image
# im is a list now, to be compat with video case
im_hf = [e[:, ::-1, :] for e in im]
# Since all frames would be same shape, just take values from 1st
im_width = im[0].shape[1]
if not cfg.MODEL.FASTER_RCNN:
box_proposals_hf = box_utils.flip_boxes(box_proposals, im_width)
else:
box_proposals_hf = None
scores_hf, boxes_hf, im_scales = im_detect_bbox(
model, im_hf, box_proposals_hf)
# Invert the detections computed on the flipped image
boxes_inv = box_utils.flip_boxes(boxes_hf, im_width)
return scores_hf, boxes_inv, im_scales
def im_detect_bbox_hflip(model,
im,
target_scale,
target_max_size,
box_proposals=None):
im_hf = im[:, ::-1, :]
im_width = im.shape[1]
if not cfg.MODEL.FASTER_RCNN:
box_proposals_hf = box_utils.flip_boxes(box_proposals, im_width)
else:
box_proposals_hf = None
scores_hf, boxes_hf, im_scale, _ = im_detect_bbox(model,
im_hf,
target_scale,
target_max_size,
boxes=box_proposals_hf)
# Invert the detections computed on the flipped image
boxes_inv = box_utils.flip_boxes(boxes_hf, im_width)
return scores_hf, boxes_inv, im_scale
def im_detect_keypoints_hflip(model, im, boxes):
"""Computes keypoint predictions on the horizontally flipped image.
Function signature is the same as for im_detect_keypoints_aug.
"""
# Compute keypoints for the flipped image
im_hf = im[:, ::-1, :]
boxes_hf = box_utils.flip_boxes(boxes, im.shape[1])
im_scales = im_conv_body_only(model, im_hf)
heatmaps_hf = im_detect_keypoints(model, im_scales, boxes_hf)
# Invert the predicted keypoints
heatmaps_inv = keypoint_utils.flip_heatmaps(heatmaps_hf)
return heatmaps_inv
def im_detect_mask_hflip(model, im, boxes):
"""Performs mask detection on the horizontally flipped image.
Function signature is the same as for im_detect_mask_aug.
"""
# Compute the masks for the flipped image
im_hf = im[:, ::-1, :]
boxes_hf = box_utils.flip_boxes(boxes, im.shape[1])
im_scales = im_conv_body_only(model, im_hf)
masks_hf = im_detect_mask(model, im_scales, boxes_hf)
# Invert the predicted soft masks
masks_inv = masks_hf[:, :, :, ::-1]
return masks_inv
def im_detect_keypoints_hflip(model, im, target_scale, target_max_size, boxes):
"""Computes keypoint predictions on the horizontally flipped image.
Function signature is the same as for im_detect_keypoints_aug.
"""
# Compute keypoints for the flipped image
im_hf = im[:, ::-1, :]
boxes_hf = box_utils.flip_boxes(boxes, im.shape[1])
im_scale = im_conv_body_only(model, im_hf, target_scale, target_max_size)
heatmaps_hf = im_detect_keypoints(model, im_scale, boxes_hf)
# Invert the predicted keypoints
heatmaps_inv = keypoint_utils.flip_heatmaps(heatmaps_hf)
return heatmaps_inv
def im_detect_mask_hflip(model, im, target_scale, target_max_size, boxes):
"""Performs mask detection on the horizontally flipped image.
Function signature is the same as for im_detect_mask_aug.
"""
# Compute the masks for the flipped image
im_hf = im[:, ::-1, :]
boxes_hf = box_utils.flip_boxes(boxes, im.shape[1])
im_scale = im_conv_body_only(model, im_hf, target_scale, target_max_size)
masks_hf = im_detect_mask(model, im_scale, boxes_hf)
# Invert the predicted soft masks
masks_inv = masks_hf[:, :, :, ::-1]
return masks_inv
def im_detect_mask_hflip(model, im, target_scale, target_max_size, boxes):
"""Performs mask detection on the horizontally flipped image.
Function signature is the same as for im_detect_mask_aug.
"""
# Compute the masks for the flipped image
im_hf = im[:, ::-1, :]
boxes_hf = box_utils.flip_boxes(boxes, im.shape[1])
# _, im_scale, _ = blob_utils.get_image_blob(im_hf, target_scale, target_max_size)
blob_conv_hf, im_scale_hf = im_conv_body_only(model, im_hf, cfg.TEST.SCALE,
cfg.TEST.MAX_SIZE)
# im_scale = im_conv_body_only(model, im_hf, target_scale, target_max_size)
masks_hf = im_detect_mask(model, im_scale_hf, boxes_hf, blob_conv_hf)
# Invert the predicted soft masks
masks_inv = masks_hf[:, :, :, ::-1]
return masks_inv
