supernotman / RetinaFace_Pytorch

Reimplement RetinaFace with Pytorch
305 stars 68 forks source link

Write eval results to tensorboard #3

Closed xsacha closed 5 years ago

xsacha commented 5 years ago

The console is the only other place the evaluation is shown and this only stays around as long as your buffer isn't full. I think it's important for this to be in the tensorboard.

-------- RetinaFace Pytorch -------- Evaluating epoch 147 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 404/404 [01:43<00:00, 3.91it/s] Recall: 0.7351636132230178 Precision: 0.9029628486077838

I've noticed my recall and precision is lower than expected but wasn't able to follow it back because it was lost in the buffer.

xsacha commented 5 years ago

Thanks