Template for converting repetitive scripts to for-loop and dict structure

repeat_structure2for_dict.py

# ref: https://stackoverflow.com/a/14241195/7961693

"""
# timestamp can be read from lidars/ folder
stamp_file_path = os.path.join(allf.lidar_dir, seq+'_stamped.npy')
lidar_stamped_dict = np.load(stamp_file_path, allow_pickle=True)
ts_np = lidar_stamped_dict.item().get('timestamp')
all_det_list = []
within_det5_list = []
within_det10_list = []
min_dist_list = []
crowd_density5_list = []
crowd_density10_list = []
for fr_idx in range(allf.nr_frames(seq_idx)):
    _, _, _, trks = allf[seq_idx, fr_idx]
    (all_det, 
    within_det5, within_det10, 
    crowd_density5, crowd_density10, 
    min_dist) = compute_metrics(trks)
    # append into list
    all_det_list.append(all_det)
    within_det5_list.append(within_det5)
    within_det10_list.append(within_det10)
    min_dist_list.append(min_dist)
    crowd_density5_list.append(crowd_density5)
    crowd_density10_list.append(crowd_density10)
ad_np = np.asarray(all_det_list, dtype=np.uint8)
wd5_np = np.asarray(within_det5_list, dtype=np.uint8)
wd10_np = np.asarray(within_det10_list, dtype=np.uint8)
md_np = np.asarray(min_dist_list, dtype=np.float32)
cd5_np = np.asarray(crowd_density5_list, dtype=np.float32)
cd10_np = np.asarray(crowd_density10_list, dtype=np.float32)
crowd_eval_dict = {'timestamp': ts_np, 
                'all_det': ad_np, 
                'within_det5': wd5_np, 
                'within_det10': wd10_np, 
                'min_dist': md_np, 
                'crowd_density5': cd5_np,
                'crowd_density10': cd10_np}
np.save(crowd_eval_npy, crowd_eval_dict)
"""

# timestamp can be read from lidars/ folder
stamp_file_path = os.path.join(allf.lidar_dir, seq+'_stamped.npy')
lidar_stamped_dict = np.load(stamp_file_path, allow_pickle=True)
ts = lidar_stamped_dict.item().get('timestamp')

# targeted metrics and correspoding dtype
attrs = ('all_det', 
        'within_det5', 'within_det10', 
        'crowd_density5', 'crowd_density10',
        'min_dist')
dtypes = (np.uint8, 
              np.uint8, np.uint8, 
              np.float32, np.float32,
              np.float32)

crowd_eval_list_dict = {k:[] for k in attrs}

for fr_idx in range(allf.nr_frames(seq_idx)):

    _, _, _, trks = allf[seq_idx, fr_idx]

    metrics = compute_metrics(trks)

    # update value for each attr
    for idx, val in enumerate(metrics):
        crowd_eval_list_dict[attrs[idx]].append(val)

crowd_eval_dict = {name:np.asarray(crowd_eval_list_dict[attrs[idx]], dtype=dtype) for idx, (name, dtype) in enumerate(zip(attrs, dtypes))}
crowd_eval_dict.update({'timestamp': ts})
np.save(crowd_eval_npy, crowd_eval_dict)