def viewResult(sample,idx,prediction,config,save_images_instead=1,size=(384,384)):
displays = []
camera_idx = 0
camera = sample['cameras'][camera_idx]
subject = sample['subject'][camera_idx]
action = sample['action'][camera_idx]
keypoints3d_pred = prediction['keypoints_3d_pred'].cpu()
keypoints_3d_pred = keypoints3d_pred[0,:, :3].detach().numpy()
keypoints_3d_gt = sample['keypoints_3d'][:, :3]
# Project and draw keypoints on images
for camera_idx in range(len(sample['cameras'])): #camera_indexes_to_show:
camera = sample['cameras'][camera_idx]
keypoints_2d_pred = project(camera.projection, keypoints_3d_pred)
keypoints_2d_gt = project(camera.projection, keypoints_3d_gt)
# import ipdb; ipdb.set_trace()
img = sample['images'][camera_idx]
pred_kind = config.pred_kind if hasattr(config, "pred_kind") else config.kind
display = vis.draw_2d_pose_cv2(keypoints_2d_pred, img, kind=pred_kind)
#display = vis.draw_2d_pose_cv2(keypoints_2d_gt, img, kind=config.kind)
cv2.putText(display, f"Cam {camera_idx:02}", (10, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display)
display3 = vis.draw_3d_pose_image(keypoints_3d_pred,kind=pred_kind,radius=450)
display3 = cv2.cvtColor(display3,cv2.COLOR_RGBA2RGB)
display3 = cv2.resize(display3, size, interpolation=cv2.INTER_AREA)
cv2.putText(display3, f"3D prediction", (10, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display3)
display3_gt = vis.draw_3d_pose_image(sample['keypoints_3d'][:, :3],kind=pred_kind,radius=450)
display3_gt = cv2.cvtColor(display3_gt,cv2.COLOR_RGBA2RGB)
display3_gt = cv2.resize(display3_gt, size, interpolation=cv2.INTER_AREA)
cv2.putText(display3_gt, f"3D GT", (10, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display3_gt)
# Fancy stacked images
for j, display in enumerate(displays):
if j == 0:
combined = display
else:
combined = np.concatenate((combined, display), axis=1)
# Load
if save_images_instead:
file = f"./result/result-{subject}-{action}-{camera.name}-{idx}.png"
cv2.imwrite(file, combined)
else:
cv2.imshow('w', combined)
cv2.setWindowTitle('w', f"Index {idx}")
c = cv2.waitKey(0) % 256
if c == ord('q') or c == 27:
print('Quitting...')
cv2.destroyAllWindows()
def viewVideoResult(sample,idx, prediction,config,size=(384,384)):
displays = []
keypoints3d_pred = prediction['keypoints_3d_pred'].cpu()
keypoints_3d_pred = keypoints3d_pred[0,:, :3].detach().numpy()
# Project and draw keypoints on images
for camera_idx in range(len(sample['cameras'])): #camera_indexes_to_show:
camera = sample['cameras'][camera_idx]
keypoints_2d_pred = project(camera.projection, keypoints_3d_pred)
# import ipdb; ipdb.set_trace()
img = sample['images'][camera_idx]
pred_kind = config.pred_kind if hasattr(config, "pred_kind") else config.kind
display = vis.draw_2d_pose_cv2(keypoints_2d_pred, img, kind=pred_kind)
cv2.putText(display, f"Cam {camera_idx:02}", (10, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display)
display3 = vis.draw_3d_pose_image(keypoints_3d_pred,kind=pred_kind,radius=450)
display3 = cv2.cvtColor(display3,cv2.COLOR_RGBA2RGB)
display3 = cv2.resize(display3, size, interpolation=cv2.INTER_AREA)
cv2.putText(display3, f"3D prediction", (10, 10), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display3)
# Fancy stacked images
for j, display in enumerate(displays):
if j == 0:
combined = display
else:
combined = np.concatenate((combined, display), axis=1)
return combined
def viewSample(sample,idx=0):
camera_idx = 0
image = sample['images'][camera_idx]
camera = sample['cameras'][camera_idx]
subject = sample['subject'][camera_idx]
action = sample['action'][camera_idx]
display = image.copy()
keypoints_2d = project(camera.projection, sample['keypoints_3d'][:, :3])
for i,(x,y) in enumerate(keypoints_2d):
cv2.circle(display, (int(x), int(y)), 3, (0,0,255), -1)
file = f"./result/{subject}-{action}-{camera.name}-{idx}.png"
cv2.imwrite(file, display)
patience = 0
sample_idx = 0 # number of sample to start from
step = 10 # in frames
while True:
sample = dataset[sample_idx]
camera_idx = 0
image = sample['images'][camera_idx]
camera = sample['cameras'][camera_idx]
display = image.copy()
from mvn.utils.multiview import project_3d_points_to_image_plane_without_distortion as project
keypoints_2d = project(camera.projection, sample['keypoints_3d'][:, :3])
for i, (x, y) in enumerate(keypoints_2d):
cv2.circle(display, (int(x), int(y)), 3, (0, 0, 255), -1)
# cv2.putText(display, str(i), (int(x)+3, int(y)), cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0,0,255))
# Get window name
sample_info = dataset.labels['table'][sample_idx]
subject_name = dataset.labels['subject_names'][sample_info['subject_idx']]
action_name = dataset.labels['action_names'][sample_info['action_idx']]
camera_name = dataset.labels['camera_names'][camera_idx]
frame_idx = sample_info['frame_idx']
cv2.imshow('w', display)
cv2.setWindowTitle(
'w', f"{subject_name}/{action_name}/{camera_name}/{frame_idx}")
print(img_dir)
camera_indexes_to_show = [0, 2, 8]
for i in range(0, len(indexes), n_images_step):
labels = dataset[i]
displays = []
# Project and draw keypoints on images
for camera_idx in range(len(labels['cameras'])): #camera_indexes_to_show:
camera = labels['cameras'][camera_idx]
keypoints_3d_pred = keypoints3d_pred[i][:, :3]
keypoints_3d_gt = labels['keypoints_3d'][:, :3]
keypoints_2d_pred = project(camera.projection, keypoints_3d_pred)
keypoints_2d_gt = project(camera.projection, keypoints_3d_gt)
# import ipdb; ipdb.set_trace()
img = labels['images'][camera_idx]
pred_kind = config.pred_kind if hasattr(config,
"pred_kind") else config.kind
display = vis.draw_2d_pose_cv2(keypoints_2d_pred, img, kind=pred_kind)
# display = vis.draw_2d_pose_cv2(keypoints_2d_gt, display, kind=config.kind)
cv2.putText(display, f"Cam {camera_idx:02}", (10, 10),
cv2.FONT_HERSHEY_SIMPLEX, 0.5, (255, 0, 0))
displays.append(display)