def gen_output(im, result, info, win_show, cour_list1, cour_list2, recist_list1, recist_list2):
im = windowing_rev(im, cfg.INPUT.WINDOWING)
im = windowing(im, win_show).astype('uint8')
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
scale = cfg.TEST.VISUALIZE.SHOW_SCALE
im = cv2.resize(im, None, None, fx=scale, fy=scale, interpolation=cv2.INTER_LINEAR)
pred = result.bbox.cpu().numpy()
labels = result.get_field('labels').cpu().numpy()
scores = result.get_field('scores').cpu().numpy()
tag_scores = result.get_field('tag_scores').cpu().numpy()
tag_predictions = result.get_field('tag_predictions').cpu().numpy()
mm2pix = info['im_scale'] / info['spacing'] * scale
contours = result.get_field('contour_mm').cpu().numpy() * mm2pix
contours = [c[c[:, 0] > 0, :] for c in contours]
contours = [c+1*scale for c in contours] # there seems to be a small offset in the mask?
recists = result.get_field('recist_mm').cpu().numpy() * mm2pix
recists += 1*scale # there seems to be a small offset in the mask?
diameters = result.get_field('diameter_mm').cpu().numpy()
pred *= scale
#contour_list = contour_list
#recist_list = recist_list
overlay, msgs = draw_results(im, pred, cour_list1, cour_list2, recist_list1, recist_list2, labels, scores, tag_predictions=tag_predictions, tag_scores=tag_scores,
contours=contours, recists=recists, diameters=diameters)
overlay = print_msg_on_img(overlay, msgs)
return overlay, msgs
def vis_all_detection(im, detections, class_names, scale):
"""
visualize all detections in one image
:param im_array: [b=1 c h w] in rgb
:param detections: [ numpy.ndarray([[x1 y1 x2 y2 score]]) for j in classes ]
:param class_names: list of names in imdb
:param scale: visualize the scaled image
:return:
"""
im = windowing_rev(im + cfg.INPUT.PIXEL_MEAN, cfg.INPUT.WINDOWING)
im = windowing(im, [-175, 275]).astype('uint8')
plt.imshow(im, cmap='gray')
for j, name in enumerate(class_names):
if name == '__background__':
continue
elif name == 'gt':
color = (0, 1, 0)
else:
color = (1, 1, 0)
# color = (random.random(), random.random(), random.random()) # generate a random color
dets = detections[j]
for det in dets:
bbox = det[:4] * scale
rect = plt.Rectangle((bbox[0], bbox[1]),
bbox[2] - bbox[0],
bbox[3] - bbox[1],
fill=False,
edgecolor=color,
linewidth=2)
plt.gca().add_patch(rect)
if det.shape[0] > 4:
score = det[-1]
plt.gca().text(
bbox[0],
bbox[1] - 2,
'{:s} {:.3f}'.format(name, score),
# bbox=dict(facecolor=color, alpha=0.5),
fontsize=12,
color=color)
plt.show()
def visualize(im, target, result, info, masker):
"""Visualize gt and predicted box, tag, mask, and RECIST"""
print('\n lesion_idxs', info['lesion_idxs'], ':', info['image_fn'])
if 'tags' in info.keys():
print('gt tags:')
for tags in target.get_field('tags'):
idx = torch.nonzero(tags > 0)
msg = ', '.join([cfg.runtime_info.tag_list[idx1] for idx1 in idx])
print(msg)
if 'diameters' in info.keys():
print('gt diameters:')
print(info['diameters'])
im = windowing_rev(im + cfg.INPUT.PIXEL_MEAN, cfg.INPUT.WINDOWING)
im = windowing(im, info['window']).astype('uint8')
im = cv2.cvtColor(im, cv2.COLOR_GRAY2RGB)
gt = target.bbox.cpu().numpy()
gt_recists = info['recists']
scale = cfg.TEST.VISUALIZE.SHOW_SCALE
im = cv2.resize(im,
None,
None,
fx=scale,
fy=scale,
interpolation=cv2.INTER_LINEAR)
gt *= scale
gt_recists *= scale
overlay, msgs = draw_results(im,
gt,
recists=gt_recists,
colors=(0, 255, 0),
thickness=(2, 1, 1))
pred = result.bbox.cpu().numpy()
labels = result.get_field('labels').cpu().numpy()
scores = result.get_field('scores').cpu().numpy()
tag_scores = result.get_field('tag_scores').cpu().numpy()
tag_predictions = result.get_field('tag_predictions').cpu().numpy()
# if cfg.MODEL.MASK_ON and len(result.get_field('contour_mm')) > 0:
mm2pix = info['im_scale'] / info['spacing'] * scale
contours = result.get_field('contour_mm').cpu().numpy() * mm2pix
contours = [c[c[:, 0] > 0, :] for c in contours]
contours = [c + 1 * scale for c in contours
] # there seems to be a small offset in the mask?
recists = result.get_field('recist_mm').cpu().numpy() * mm2pix
recists += 1 * scale # there seems to be a small offset in the mask?
diameters = result.get_field('diameter_mm').cpu().numpy()
pred *= scale
overlay, msgs = draw_results(overlay,
pred,
labels,
scores,
tag_predictions=tag_predictions,
tag_scores=tag_scores,
contours=contours,
recists=recists,
diameters=diameters)
plt.figure(1)
plt.imshow(overlay)
for msg in msgs:
print(msg)
if cfg.TEST.VISUALIZE.SHOW_MASK_HEATMAPS and 'mask' in result.extra_fields.keys(
):
plt.figure(2)
mask = result.get_field('mask')
mask = masker([mask], [result])[0]
mask = mask.numpy()
mask0 = cv2.resize(mask[0, 0], None, None, fx=scale, fy=scale)
mask1 = np.empty(
(mask.shape[0], mask.shape[1], mask0.shape[0], mask0.shape[1]),
dtype=mask.dtype)
for i1 in range(mask.shape[0]):
for i2 in range(mask.shape[1]):
mask1[i1, i2] = cv2.resize(mask[i1, i2],
None,
None,
fx=scale,
fy=scale,
interpolation=cv2.INTER_NEAREST)
pred *= scale
heatmap = np.sum(mask1, (0, 1)).astype('uint8')
# heatmap = cv2.applyColorMap(heatmap, cv2.COLORMAP_JET)
plt.imshow(heatmap, cmap='gray')
plt.show()