generate_json_mask.py not generating absolute path in output filelist and json file

[Akimoto-Cris]

After executing the following script:

python generate_json_mask.py \
    --ann_path ~/datasets/COCO2017/annotations/person_keypoints_val2017.json \
    --json_path ~/datasets/COCO2017/results/json/val2017.json \
    --masklist_path ~/datasets/COCO2017/results/masklist/val2017.txt \
    --filelist_path ~/datasets/COCO2017/results/filelist/val2017.txt \
    --mask_dir ~/datasets/COCO2017/results/masks/mask_val

The generated filelist and filenames in json files should be absolute path, but i got these:

In filelist.txt

000000472209.jpg
000000019642.jpg
000000268909.jpg
000000547777.jpg
000000049429.jpg
000000375321.jpg
000000250249.jpg
000000545549.jpg
000000126073.jpg
000000374391.jpg

In xxx.json:

[{"filename": "000000391895.jpg", "info": [{"pos": [416.82, 172.525], "keypoints": [[
368, 61, 0], [401.5, 82.5, 1], [435, 81, 1], [446, 125, 1], [0, 0, 2], [368, 84, 1], 
[362, 125, 1], [360, 153, 1], [439, 166, 0], [461, 234, 1], [474, 287, 1], [397, 167,
 0], [369, 193, 1], [361, 246, 1], [0, 0, 2], [369, 52, 1], [0, 0, 2], [382, 48, 1]],
 "scale": 0.8172010869565218}]}, 
...

I failed to run train_pose.py, maybe caused by this problem. I did prepared annotations for training set as well, using similar script.

Did I run the script in wrong directory? Please tell me how to fix it, other than rewriting the python script.