northeastsquare opened a new issue #17695: mxnet inference wrong
URL: https://github.com/apache/incubator-mxnet/issues/17695
 
 
   ## Description
   I train mask classification, validation acc:98%, but inference with train 
and val image, many case is wrong, maybe 25% is wrong
   the codes in:https://github.com/northeastsquare/mxnet_mask_classification
   ## Occurrences
   in training:
   ![Screenshot from 2020-02-26 
22-56-56](https://user-images.githubusercontent.com/3086078/75356965-98108600-58eb-11ea-94e5-95334ed85caf.png)
   but inference:
   
![image](https://user-images.githubusercontent.com/3086078/75357058-bb3b3580-58eb-11ea-9523-58bf3f79ab9d.png)
   because all the image's class is 1, so the result is not right
   ## What have you tried to solve it?
   
   1. use this to save checkpoint: export_block('mobilenetv2-0.5-new', 
net,preprocess=True, ctx=mx.gpu(0)), and inference no need  to subtract mean, 
divide std.
   2. use inference.py to inference the result, but are wrong
   

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to