def extract_image_patches(image, bboxes):
bboxes = np.round(bboxes).astype(np.int)
bboxes = bbox_utils.clip_boxes(bboxes, image.shape)
patches = [
img_process(image[box[1]:box[3], box[0]:box[2]]) for box in bboxes
]
return np.array(patches)
def decoder(self,
ims,
anchors,
cls_score,
bbox_pred,
thresh=0.6,
nms_thresh=0.2,
test_conf=None):
if test_conf is not None:
thresh = test_conf
bboxes = self.box_coder.decode(anchors, bbox_pred, mode='xywht')
bboxes = clip_boxes(bboxes, ims)
scores = torch.max(cls_score, dim=2, keepdim=True)[0]
keep = (scores >= thresh)[0, :, 0]
if keep.sum() == 0:
return [torch.zeros(1), torch.zeros(1), torch.zeros(1, 5)]
scores = scores[:, keep, :]
anchors = anchors[:, keep, :]
cls_score = cls_score[:, keep, :]
bboxes = bboxes[:, keep, :]
# NMS
anchors_nms_idx = nms(
torch.cat([bboxes, scores], dim=2)[0, :, :], nms_thresh)
nms_scores, nms_class = cls_score[0, anchors_nms_idx, :].max(dim=1)
output_boxes = torch.cat(
[bboxes[0, anchors_nms_idx, :], anchors[0, anchors_nms_idx, :]],
dim=1)
return [nms_scores, nms_class, output_boxes]
def predict(self, rois):
"""
:param rois: numpy array [N, 4] ( x1, y1, x2, y2)
:return: scores [N]
"""
scaled_rois = rois * self.im_scale
cls_scores = self.model.get_cls_score_numpy(self.score_map, scaled_rois)
# check area
rois = rois.reshape(-1, 4)
clipped_boxes = bbox_utils.clip_boxes(rois, self.ori_image_shape)
ori_areas = (rois[:, 2] - rois[:, 0]) * (rois[:, 3] - rois[:, 1])
areas = (clipped_boxes[:, 2] - clipped_boxes[:, 0]) * (clipped_boxes[:, 3] - clipped_boxes[:, 1])
ratios = areas / np.clip(ori_areas, a_min=1e-4, a_max=None)
cls_scores[ratios < 0.5] = 0
return cls_scores
def im_detect_bbox_aug(net, im, scales, ctx, bbox_stds=(1, 1, 1, 1),
hflip=False, vflip=False, vhflip=False,
threshold=1e-3, nms_threshold=.3,
viz=False, pad=False, class_agnostic=True):
all_bboxes = []
all_scores = []
img_ori = im.copy()
for scale_min, scale_max in scales:
fscale = min(1.0 * scale_min / min(img_ori.shape[:2]), 1.0 * scale_max / max(img_ori.shape[:2]))
img_resized = cv2.resize(img_ori, (0, 0), fx=fscale, fy=fscale, interpolation=cv2.INTER_CUBIC)
h, w, c = img_resized.shape
if pad:
h_padded = h if h % 32 == 0 else h + 32 - h % 32
w_padded = w if w % 32 == 0 else w + 32 - w % 32
else:
h_padded = h
w_padded = w
img_padded = np.zeros(shape=(h_padded, w_padded, c), dtype=img_resized.dtype)
img_padded[:h, :w, :] = img_resized
data = nd.array(img_padded.transpose(2, 0, 1)[np.newaxis], ctx=ctx[0])
im_info = nd.array([[h, w, 1.0]], ctx=ctx[0])
rois, scores, bbox_deltas = net(data, im_info)
rois = rois[:, 1:]
if isinstance(rois, mx.nd.NDArray):
rois = rois.asnumpy()
bbox_deltas = bbox_deltas[0].asnumpy()
bbox_deltas = pre_compute_deltas(bbox_deltas, bbox_stds=bbox_stds)
bbox = bbox_pred(rois, bbox_deltas)
bbox = clip_boxes(bbox, data.shape[2:4])
bbox /= fscale
all_bboxes.append(bbox)
all_scores.append(scores[0].asnumpy())
# flip
if hflip:
rois, scores, bbox_deltas = net(data[:, :, :, ::-1], im_info)
if isinstance(rois, mx.nd.NDArray):
rois = rois.asnumpy()
rois = rois[:, 1:]
bbox_deltas = bbox_deltas[0].asnumpy()
bbox_deltas = pre_compute_deltas(bbox_deltas, bbox_stds=bbox_stds)
bbox = bbox_pred(rois, bbox_deltas)
bbox = clip_boxes(bbox, data.shape[2:4])
tmp = bbox[:, 0::4].copy()
bbox[:, 0::4] = data.shape[3] - bbox[:, 2::4] # x0 = w - x0
bbox[:, 2::4] = data.shape[3] - tmp # x1 = w -x1
bbox /= fscale
all_bboxes.append(bbox)
all_scores.append(scores[0].asnumpy())
if vflip:
rois, scores, bbox_deltas = net(data[:, :, ::-1, :], im_info)
if isinstance(rois, mx.nd.NDArray):
rois = rois.asnumpy()
rois = rois[:, 1:]
bbox_deltas = bbox_deltas[0].asnumpy()
bbox_deltas = pre_compute_deltas(bbox_deltas, bbox_stds=bbox_stds)
bbox = bbox_pred(rois, bbox_deltas)
bbox = clip_boxes(bbox, data.shape[2:4])
tmp = bbox[:, 1::4].copy()
bbox[:, 1::4] = data.shape[2] - bbox[:, 3::4] # y0 = h - y1
bbox[:, 3::4] = data.shape[2] - tmp # y1 = h -y0
bbox /= fscale
all_bboxes.append(bbox)
all_scores.append(scores[0].asnumpy())
if vhflip:
rois, scores, bbox_deltas = net(data[:, :, ::-1, ::-1], im_info)
if isinstance(rois, mx.nd.NDArray):
rois = rois.asnumpy()
rois = rois[:, 1:]
bbox_deltas = bbox_deltas[0].asnumpy()
bbox_deltas = pre_compute_deltas(bbox_deltas, bbox_stds=bbox_stds)
bbox = bbox_pred(rois, bbox_deltas)
bbox = clip_boxes(bbox, data.shape[2:4])
tmp = bbox[:, 0::4].copy()
bbox[:, 0::4] = data.shape[3] - bbox[:, 2::4] # x0 = w - x0
bbox[:, 2::4] = data.shape[3] - tmp # x1 = w -x1
tmp = bbox[:, 1::4].copy()
bbox[:, 1::4] = data.shape[2] - bbox[:, 3::4] # y0 = h - y1
bbox[:, 3::4] = data.shape[2] - tmp # y1 = h -y0
bbox /= fscale
all_bboxes.append(bbox)
all_scores.append(scores[0].asnumpy())
all_bboxes = np.concatenate(all_bboxes, axis=0)
all_scores = np.concatenate(all_scores, axis=0)
pred_bboxes = []
pred_scores = []
pred_clsid = []
mx.nd.waitall()
from utils.nms.nms import gpu_nms_wrapper
nms_wrapper = gpu_nms_wrapper(thresh=.3, device_id=7)
for j in range(1, all_scores.shape[1]):
cls_scores = all_scores[:, j, np.newaxis]
cls_boxes = all_bboxes[:, 4:8] if class_agnostic else all_bboxes[:, j * 4:(j + 1) * 4]
cls_dets = np.hstack((cls_boxes, cls_scores))
# cls_dets = nd.contrib.box_nms(nd.array(cls_dets, ctx=mx.cpu(ctx[0].device_id)),
# overlap_thresh=nms_threshold, coord_start=0, score_index=4, id_index=-1,
# force_suppress=True, in_format='corner',
# out_format='corner').asnumpy()
keep = nms_wrapper(cls_dets.astype('f'))
cls_dets = cls_dets[keep]
cls_dets = cls_dets[cls_dets[:, -1] > threshold, :]
pred_bboxes.append(cls_dets[:, :4])
pred_scores.append(cls_dets[:, 4])
pred_clsid.append(j * np.ones(shape=(cls_dets.shape[0],), dtype=np.int))
pred_bboxes = np.concatenate(pred_bboxes, axis=0)
pred_scores = np.concatenate(pred_scores, axis=0)
pred_clsid = np.concatenate(pred_clsid, axis=0)
if viz:
import gluoncv
import matplotlib.pyplot as plt
gluoncv.utils.viz.plot_bbox(img_ori, bboxes=pred_bboxes, scores=pred_scores, labels=pred_clsid,
thresh=.5)
plt.show()
return pred_bboxes, pred_scores, pred_clsid