def run_eval(image_dir, anno_dir, vis_dir, image_list_txt, model, preprocess):
"""Run the evaluation on the test set and report mAP score
:param model: the model to test
:returns: float, the reported mAP score
"""
# This txt file is fount in the caffe_rtpose repository:
# https://github.com/CMU-Perceptual-Computing-Lab/caffe_rtpose/blob/master
img_ids, img_paths, img_heights, img_widths = get_coco_val(
image_list_txt)
# img_ids = img_ids[81:82]
# img_paths = img_paths[81:82]
print("Total number of validation images {}".format(len(img_ids)))
# iterate all val images
outputs = []
print("Processing Images in validation set")
for i in range(len(img_ids)):
if i % 10 == 0 and i != 0:
print("Processed {} images".format(i))
oriImg = cv2.imread(os.path.join(image_dir, 'val2014/' + img_paths[i]))
# Get the shortest side of the image (either height or width)
shape_dst = np.min(oriImg.shape[0:2])
# Get results of original image
multiplier = get_multiplier(oriImg)
orig_paf, orig_heat = get_outputs(
multiplier, oriImg, model, preprocess)
# Get results of flipped image
swapped_img = oriImg[:, ::-1, :]
flipped_paf, flipped_heat = get_outputs(multiplier, swapped_img,
model, preprocess)
# compute averaged heatmap and paf
paf, heatmap = handle_paf_and_heat(
orig_heat, flipped_heat, orig_paf, flipped_paf)
# choose which post-processing to use, our_post_processing
# got slightly higher AP but is slow.
param = {'thre1': 0.1, 'thre2': 0.05, 'thre3': 0.5}
canvas, to_plot, candidate, subset = decode_pose(
oriImg, param, heatmap, paf)
vis_path = os.path.join(vis_dir, img_paths[i])
cv2.imwrite(vis_path, to_plot)
# subset indicated how many peoples foun in this image.
append_result(img_ids[i], subset, candidate, outputs)
# cv2.imshow('test', canvas)
# cv2.waitKey(0)
# Eval and show the final result!
return eval_coco(outputs=outputs, dataDir=anno_dir, imgIds=img_ids)
def viz_keypoints(oriImg, p_heatmap, p_paf, name):
# multiplier = get_multiplier(oriImg, 368.)
# scale_search = [1.]
# multiplier = [x * 481. / float(oriImg.shape[0]) for x in scale_search]
multiplier = [1.]
paf, heatmap = _get_outputs(multiplier, oriImg, p_heatmap, p_paf, numkeypoints=NUM_KEYPOINTS, numlims=NUM_LIMBS)
# Given a (grayscale) image, find local maxima whose value is above a given threshold (param['thre1'])
# Criterion 1: At least 80% of the intermediate points have a score higher than thre2
param = {'thre1': 0.1, 'thre2': 0.05, 'thre3': 0.5}
to_plot, canvas, joint_list, person_to_joint_assoc = decode_pose(oriImg, param, heatmap, paf)
cv2.imwrite(name, canvas)
def extract_pose(video, oriImg, name, filestream, model):
# Get results of original image
multiplier = get_multiplier(oriImg)
with torch.no_grad():
paf, heatmap = get_outputs(multiplier, oriImg, model, 'rtpose')
pose_cords = compute_cordinates(heatmap, paf, oriImg)
# coordinate
print("{}: {}: {}".format(
str(name) + ".jpg", str(list(pose_cords[:, 0])),
str(list(pose_cords[:, 1]))),
file=filestream)
filestream.flush()
canvas, to_plot, candidate, subset = decode_pose(oriImg, param, heatmap,
paf)
save_result(video, oriImg, to_plot, name)
def process(model, oriImg, process_speed):
# Get results of original image
multiplier = get_multiplier(oriImg, process_speed)
with torch.no_grad():
orig_paf, orig_heat = get_outputs(multiplier, oriImg, model, 'rtpose')
# Get results of flipped image
swapped_img = oriImg[:, ::-1, :]
flipped_paf, flipped_heat = get_outputs(multiplier, swapped_img, model,
'rtpose')
# compute averaged heatmap and paf
paf, heatmap = handle_paf_and_heat(orig_heat, flipped_heat, orig_paf,
flipped_paf)
param = {'thre1': 0.1, 'thre2': 0.05, 'thre3': 0.5}
to_plot, canvas, joint_list, person_to_joint_assoc = decode_pose(
oriImg, param, heatmap, paf)
return to_plot, canvas, joint_list, person_to_joint_assoc
model = get_model(trunk='vgg19')
model = torch.nn.DataParallel(model).cuda()
model.load_state_dict(torch.load(weight_name))
model.float()
model.eval()
model = model.cuda()
test_image = 'kids.jpg'
oriImg = cv2.imread(test_image) # B,G,R order
shape_dst = np.min(oriImg.shape[0:2])
# Get results of original image
multiplier = get_multiplier(oriImg)
with torch.no_grad():
orig_paf, orig_heat = get_outputs(multiplier, oriImg, model, 'rtpose')
# Get results of flipped image
swapped_img = oriImg[:, ::-1, :]
flipped_paf, flipped_heat = get_outputs(multiplier, swapped_img, model,
'rtpose')
# compute averaged heatmap and paf
paf, heatmap = handle_paf_and_heat(orig_heat, flipped_heat, orig_paf,
flipped_paf)
param = {'thre1': 0.1, 'thre2': 0.05, 'thre3': 0.5}
canvas, to_plot, candidate, subset = decode_pose(oriImg, param, heatmap, paf)
cv2.imwrite('result_photo.png', to_plot)
model.eval()
for i, fname in enumerate(loader):
oriImg = cv2.imread(fname) # B,G,R order
shape_dst = np.min(oriImg.shape[0:2])
# Get results of original image
multiplier = get_multiplier(oriImg)
with torch.no_grad():
orig_paf, orig_heat = get_outputs(multiplier, oriImg, model, 'rtpose')
# Get results of flipped image
swapped_img = oriImg[:, ::-1, :]
flipped_paf, flipped_heat = get_outputs(multiplier, swapped_img, model,
'rtpose')
# compute averaged heatmap and paf
paf, heatmap = handle_paf_and_heat(orig_heat, flipped_heat, orig_paf,
flipped_paf)
param = {'thre1': 0.1, 'thre2': 0.05, 'thre3': 0.5}
canvas, to_plot, joint_list, person_to_joint_assoc = decode_pose(
oriImg, param, heatmap, paf)
saver.crawl(fname, joint_list, person_to_joint_assoc)
cv2.imwrite(dir + 'done\\' + str(i) + '.png', to_plot)
print('%d images have been annotated!' % i)
print('Annotation completed!')
saver.distribute()
exit()
