Open chriszeng87 opened 6 years ago
@chriszeng87 yes, I'm working in this now
Cool!
Why I downloaded the data set is not the same with you?My json file name is keypoint_train_annottaions_20170909.json, json file is not high and wide, do you have any work on the json file?
@zoukai214 I think the author modified the json files. in main.py line:103 m = np.zeros([human_nums, annotations['width'], annotations['height'], 14]) in numpy array, it should be m = np.zeros([human_nums, annotations['height'], annotations['width'], 14]) . So @RodrigoGantier did you write origin image width to annotations['height'], and height to annotations['width']?
@nebuladream I think the author modified the json files in main.py line:89
"masks: A bool array of shape [height, width, instance count] withone mask per instance"
height and width meaning is bounding box height and width, use json file "human_annotations" calculate.Do not know if this understanding is correct
the modified json files are updated in released part sorry for not mention, I forgot
@nebuladream I dont remember XD
Could you share the AI challenger dataset you are using? The AI challenger website reject new registration...
@RodrigoGantier why download the modified json files, but some picture can't found
@RodrigoGantier
I run main.py and the dataset is person keypoints of COCO 2017(annotations is 'person_keypoints_train2017.json') , when the code run to the 64th linejson_file = pd.read_json(os.path.join(data_dir, json_path_dict[subset]))
,I got the error
ValueError: arrays must all be same length
I know use coco API to deal with 'person_keypoints_train2017.json',but for pandas.read_json, I want to know how do you use it ?
Thank you for sharing this awesome project. Have you used coco 2017 keypoints data to train?