def create_dataframe_faces():
# 取读所有的mask图，构成一个dataframe
faces = pd.DataFrame([])
i = 0
arg = setting.args_class()
for video_path,json_path,contour_path in zip(arg.videodir,arg.tracked_json,arg.contdir):
img_dict = contour_path
data = contour_utils.load_json(json_path)
for img_name in tqdm(os.listdir(img_dict)):
if not(img_name[-1] == 'g'):
print(img_name+' is not used')
continue
if np.random.randn()>0.8:
print('add %d th image '%(len(faces+1)))
# print(i,img_dict + img_name)
frame = plt.imread(img_dict + img_name)
frame = np.asarray(frame[:,:,0],dtype = 'uint8')
frame[frame == 255] = 1
# flat = frame.flatten()
# print(flat.shape)
face = pd.Series(frame.flatten(),name = img_name)
# print(face)
faces = faces.append(face)
i += 1
if len(faces)>2500:
break
if len(faces)>2500:
break
# ## 可视化
# width, height = frame.shape
# fig, axes = plt.subplots(10,10,figsize=(9,9),
# subplot_kw = {'xticks':[], 'yticks':[]},
# gridspec_kw = dict(hspace=0.01, wspace=0.01))
# for i, ax in enumerate(axes.flat):
# ax.imshow(faces.iloc[i].values.reshape(height,width),cmap='gray')
# plt.savefig('love.png')
import pickle
print('writing file...')
with open('./faces.pckl', 'wb') as f:
pickle.dump(faces, f)
def pca_for_faces(show_fig=False):
arg = setting.args_class()
print('reading...')
with open('./faces.pckl', 'rb') as f:
faces = pickle.load(f)
#n_components=0.80 means it will return the Eigenvectors that have the 80% of the variation in the dataset
print('calculating...')
faces_pca = PCA(n_components=0.9)
faces_pca.fit(faces)
if show_fig:
fig, axes = plt.subplots(2,
5,
figsize=(9, 3),
subplot_kw={
'xticks': [],
'yticks': []
},
gridspec_kw=dict(hspace=0.01, wspace=0.01))
for i, ax in enumerate(axes.flat):
ax.imshow(faces_pca.components_[i].reshape(200, 200), cmap='gray')
plt.savefig('love_pc.png')
plt.close()
plt.figure()
plt.plot(np.cumsum(faces_pca.explained_variance_ratio_))
plt.xlabel('number of components')
plt.ylabel('cumulative explained variance')
plt.savefig('love_variance.png')
plt.close()
print('writing...')
with open('./pca.pckl', 'wb') as f:
pickle.dump(faces_pca, f)
print(faces)
components = faces_pca.transform(faces)
count_used += 1
continue
read_flag, frame = cap.read()
i += 1
cap.release()
print(
'%d frames dont have 2 mice, %d frames have points outside pic. %d frame will be used, %d frame in total'
% (count_lessMouse, count_out, count_used, count_frame))
if __name__ == '__main__':
# dir_name = '/disk1/zexin/project/mice/clustering_sequencial/forZexin/results/0603/1411_black_two/'
# video_name = '1411_black_two.mov'
# json_name= '/disk1/zexin/project/mice/clustering_sequencial/forZexin/results/0603/1411_black_two/alphapose-results-forvis-tracked.json'
arg = setting.args_class()
if (len(sys.argv) != 1):
all_video_paths_id = [int(sys.argv[1])]
print('only deal with the %d th video path: \n ' % (int(sys.argv[1])))
else:
all_video_paths_id = [i for i in range(len(arg.videodir))]
for i in all_video_paths_id:
video_path = arg.videodir[i]
json_path = arg.tracked_json[i]
contour_path = arg.contdir[i]
get_samples(video_path, json_path, contour_path, arg, targetMouseID=0)
get_samples(video_path, json_path, contour_path, arg, targetMouseID=1)