Open spoonbobo opened 2 years ago
Hi, any folks have ideas how to solve this error? Thanks.
Could you please check if image_size
and heatmap_size
are properly set in the ann_info
in the data samples? Please check following references:
Hi, any folks have ideas how to solve this error? Thanks.
Hi folks, I used MMPose to train on CarFusion dataset using HRNet with UDP implemented (hrnet_w32_coco_384x288_udp). The error
ValueError: cannot convert float NaN to integer
happened as I passed the following code in the training validation pipeline.dict(type='TopDownAffine', use_udp=True)
Without use_udp=True (False), the training is working fine.Training config:
Environment
Error Traceback