fangyh09/posedatasets

Filter multiple pose datasets (coco, flic, lsp, mpii, ai_challenge)

PoseDatasets

Usage

python ai_dataset.py
python coco_dataset.py
python mpii_dataset.py
python lsp_dataset.py

Default setting

# KEYPOINT_NUM = 10
# WIDTH_RATIO = 0.2
# HEIGHT_RATIO = 0.3
# PERSON_NUM = 10
filter_configs = {
    "c1": [10, 10, 0.2, 0.3],
    "c2": [10, 13, 0.2, 0.3],
    "c3": [5, 10, 0.2, 0.3],
    "c4": [5, 8, 0.2, 0.3],
    "c5": [8, 10, 0.05, 0.1],
    "c6": [8, 8, 0.1, 0.15],
    "c7": [8, 8, 0.3, 0.6]
}
mode = "c7"

You can use different mode or self-designed mode in img_filter.py

MPII data format

AI challenge data format

Attention⚠️

when judge vis > 0

posedatasets's Issues

mpii data？

the mpii data dont have the test dataset???
I must split it from the train dataset??? and use 0.9 0.1 to split it??

TypeError: '>' not supported between instances of 'dict_values' and 'int‘

python mpii_dataset.py have error as：
mpii_dataset.py:79: FutureWarning: elementwise != comparison failed and returning scalar instead; this will raise an error or perform elementwise comparison in the future.
if annopoint != []:
mpii_dataset.py:102: DeprecationWarning: The truth value of an empty array is ambiguous. Returning False, but in future this will result in an error. Use array.size > 0 to check that an array is not empty.
for v in annopoint['is_visible'][0]]
Traceback (most recent call last):
File "mpii_dataset.py", line 192, in
save_joints()
File "mpii_dataset.py", line 122, in save_joints
if not ok(feed_dict):
File "/content/drive/MyDrive/HumanPoseDemo/clone/PoseDatasets/img_filter.py", line 56, in ok
keypoint_num = np.sum(vis > 0)
TypeError: '>' not supported between instances of 'dict_values' and 'int'
How to solve？please！

Could you give me the AI Challenge pose estimation data?

Hi,buddy! I am a student in CHINA.Could you give me the AI Challenge pose estimation data? I've been looking for it for a long time and can't be downloaded on the official website.Many thanks!