data2Mat.py

import json
import numpy as np
import scipy.io as sio
WIDTH=960
HEIGHT=540


camera_space_point_type = np.dtype({'names':['x', 'y', 'z'],'formats':['f','f', 'f']}, align= True )
filename='app/static/result/haha'
savefile='shirt_1'
savefile='newspaper_2'
savefile='post_bjtu'
savefile='post_olm'
savefile='shirt_1'
savefile='bjtu'
DEPTH_PATH = "app/static/data/"

with open(filename) as f:
    data = json.load(f)
kvs = data['videos']['app/static/data/'+savefile]['data'].items()
kvss = sorted(kvs)
data_lst = list(map(lambda x:list(zip(x[1]['mark_list'],x[1]['coordinate_list'])),kvss))

st_f=0
# st_f=112
ed_f=len(kvss)
# ed_f=216
def cal_difference_value(index,w,depth):
    for i in range(int(w)):
        idx1 = index - i
        idx2 = index + i
        if not (np.inf in depth[idx2] or -np.inf in depth[idx2]) and \
                not (np.inf in depth[idx1] or -np.inf in depth[idx1]):
            ret_lst = []
            for a, b in zip(depth[idx1], depth[idx2]):
                ret_lst.append((float(a) + float(b)) / 2)
            return ret_lst

lst=[]
#st_f=20
for imgs in range(st_f,ed_f):
    if kvss[imgs][0].startswith('8'):
        continue
    t_lst = []
    depth = np.fromfile(DEPTH_PATH + savefile + '/depth/' + kvss[imgs][0].split('.')[0] + '.txt',dtype=camera_space_point_type)
    print(imgs, kvss[imgs][0])
    for point in range(len(data_lst[imgs])):
        loc = WIDTH * int(data_lst[imgs][point][0]['y']) + int(data_lst[imgs][point][0]['x'])
        if np.inf in depth[loc] or -np.inf in depth[loc]:
            print('inf,',imgs,point,depth[loc])
            ret_lst = cal_difference_value(loc,int(data_lst[imgs][point][0]['x']),depth)
            t_lst.append([int(data_lst[imgs][point][0]['x']),int(data_lst[imgs][point][0]['y']),ret_lst[0],ret_lst[1],ret_lst[2]])
        else:
            t_lst.append([int(data_lst[imgs][point][0]['x']),int(data_lst[imgs][point][0]['y']),depth[loc][0],depth[loc][1],depth[loc][2]])
    lst.append(t_lst)
save_data = {}
save_data['data'] = lst
save_data['file'] = list(map(lambda x:x[0], kvss))[st_f:ed_f]

sio.savemat(savefile+'2.mat',save_data)