def FAN_model(resource, model):
preds = fan_labeler(resource)
# create structure for AI Labeler result
frames_name = []
frames_kpts = []
types = ('*.jpg', '*.png')
files_grabbed = []
for files in types:
files_grabbed.extend(glob.glob(os.path.join(resource, files)))
im_list = sorted(files_grabbed)
for i in range(len(preds)):
sp_kpts = []
frames_name.append(im_list[i])
arr = np.empty([2, 68])
for j in range(len(preds[0])):
arr[0, j] = preds[i][j][0]
arr[1, j] = preds[i][j][1]
sp_kpts.append(arr)
frames_kpts.append(sp_kpts)
predictions = {}
predictions['all_keyps'] = [[], frames_kpts]
predictions['all_boxes'] = [[] for i in range(len(frames_kpts))]
helpers.savepkl(predictions, resource, model)
def update_flags(self):
global result
for num in range(num_poses):
result['all_keyps'][1][self.idx][num] = np.append(
result['all_keyps'][1][self.idx][num],
np.ones((1, num_kpts)),
axis=0)
result['all_boxes'][self.idx].append(bbox)
for i in range(len(fixed)):
pose = int(fix[i] / num_kpts)
joint = fix[i] % num_kpts
# print(result['all_keyps'][1][self.idx][pose][0][joint])
result['all_keyps'][1][self.idx][pose][0][joint] = fixed[i][0]
result['all_keyps'][1][self.idx][pose][1][joint] = fixed[i][1]
result['all_keyps'][1][self.idx][pose][-1][joint] = fixed[i][2]
self.idx = self.idx + 1
if self.idx < len(self.frames_kpts):
self.show_flags()
else:
messagebox.showinfo(
"Information",
"All frames are revised and keypoints are saved!")
helpers.savepkl(result, self.resource, "gt")
plt.close()
def detectron2_model(resource, model):
preds = inference(resource)
# create structure for AI Labeler result
frames_name = []
frames_kpts = []
types = ('*.jpg', '*.png')
files_grabbed = []
for files in types:
files_grabbed.extend(glob.glob(os.path.join(resource, files)))
im_list = sorted(files_grabbed)
for i in range(len(preds)):
sp_kpts = []
frames_name.append(im_list[i])
arr = np.empty([2, 17])
filename = os.path.basename(im_list[i])
print(filename)
print(len(preds[filename]['keypoints']))
if len(preds[filename]['keypoints']) == 0:
os.remove(os.path.join(resource, filename))
continue
for j in range(len(preds[filename]['keypoints'][0])):
arr[0, j] = preds[filename]['keypoints'][0][j][0]
arr[1, j] = preds[filename]['keypoints'][0][j][1]
sp_kpts.append(arr)
frames_kpts.append(sp_kpts)
predictions = {}
predictions['all_keyps'] = [[], frames_kpts]
predictions['all_boxes'] = [[] for i in range(len(frames_kpts))]
helpers.savepkl(predictions, resource, model)
def update_figure(self):
global flag
list_flag = []
list = [1] * num_kpts * num_poses
for i in range(len(vis_idx)):
list[vis_idx[i]] = flag[i]
for pose in range(num_poses):
list_flag.append(list[pose * num_kpts:(pose + 1) * num_kpts])
dict_flags[str(self.idx)] = list_flag
print(dict_flags)
self.idx = self.idx + 1
if self.idx < len(self.frames_kpts):
self.show_figure()
else:
messagebox.showinfo(
"Information", "All frames are reviewed and flags are saved!")
helpers.savepkl(dict_flags, self.resource, "flag")
plt.close()
def hourglass_model(resource, model):
preds = hg_labeler(resource)
# create structure for AI Labeler result
frames_name = []
frames_kpts = []
im_list = sorted(glob.glob(os.path.join(resource, "*.jpg")))
for i in range(len(preds)):
sp_kpts = []
frames_name.append(im_list[i])
arr = np.empty([2, 16])
for j in range(len(preds[0])):
arr[0, j] = preds[i][j][0]
arr[1, j] = preds[i][j][1]
sp_kpts.append(arr)
frames_kpts.append(sp_kpts)
predictions = {}
predictions['all_keyps'] = [[], frames_kpts]
predictions['all_boxes'] = [[] for i in range(len(frames_kpts))]
helpers.savepkl(predictions, resource, model)