hsp2454

I want to generate the image skeleton in the test set, isn't it "--out save path", every time I want to save the test results error 

Hello, what is the general reason for this situation? When I was training in mmpose0.x before, when I tested demo.py, I could completely identify the key points.

Thank you for your answer. When I changed the batch_size in val_dataloader from 32 to 2 and ran test.py, it reported the following error.  But when I evaluate it...

Also when I run demo\image_demo.py, it only detects a few points. I think my data is fine, I have trained HRNet in mmpose0.x, it can identify all the points

This is the information in my D:\Code\mmpose-1.x\configs\ base \datasets\pig.py，And no changes have been made to the information in my mmpose\evaluation\metrics\coco_metric.py file ``` dataset_info = dict( dataset_name='PigCocoDataset', paper_info=dict( author='Lin, Tsung-Yi and...

sorry i misunderstood ``` val_evaluator = [ dict(type='PCKAccuracy', thr=0.2), dict(type='AUC'), dict(type='EPE'), ] ``` Yes, my custom data set is indeed 21 points

This is my dataset format, I copied other datasets directly, only changed the data path, and the detected APs are all 0 ``` # Copyright (c) OpenMMLab. All rights reserved....

I switched to 0.x to use it later, but it still doesn't work, because when I first trained hrnet with 0.x, I used PCK for evaluation, but now I want...

Sorry, 0.x does not have this configuration, all my configurations are on it, and the configuration file only has this `evaluation = dict(interval=1, metric='mAP', save_best='AP')`

Here is my config file in 0.x ``` Config: checkpoint_config = dict(interval=2) log_config = dict( interval=1, hooks=[dict(type='TextLoggerHook'), dict(type='TensorboardLoggerHook')]) log_level = 'INFO' load_from = None resume_from = None dist_params = dict(backend='nccl')...