search

yuto3o / yolox

More Than YOLO（v3, v4, v3-tiny, v4-tiny）
150 stars 46 forks source link

IndexError: index 3 is out of bounds for axis 0 with size 3 #14

Closed amirhosseinghdv closed 3 years ago

amirhosseinghdv commented 3 years ago

@yuto3o Thank you for your great code. I did everything as explained, but while training i keep receiving the error: "IndexError: index 3 is out of bounds for axis 0 with size 3" after some train steps. what do you think could be the problem? Thanks.

yuto3o commented 3 years ago

Maybe there are some mistakes in the data.

amirhosseinghdv commented 3 years ago

@yuto3o Thankfully i could manage to find the issue with my data, but unfortunately now while training, as mentioned by other users, for all epochs i get the massage "AP not improved from 0.00%" and the best loss is 12. it happens whether using pre-trained weights or not. i could not find the solution by reading those discussions. i really appreciate it if you could help.

yuto3o commented 3 years ago

It is not a simple question. Instead of using default config, please try more hyperparameters.

amirhosseinghdv commented 3 years ago

@yuto3o Thanks again, i changed some hyperparameters and could reach AP of 35%, but i had reached AP of 60% using Alexey AB repository by changing it's config according to it's guiding based on number of classes with the same amount of data, but your code's configs file is different and i wasn't able to do so. would you please provide some help on how to set the hyperparameters according to our number of classes in order to achieve the best result with your code. Thanks

yuto3o commented 3 years ago

Yep, this repo does not guarantee the same behavior as Alexey's repo do. Maybe you can use hyperparameter searching toolkit like https://github.com/scikit-optimize/scikit-optimize.