Number of batches remain same after augmentation

experiencor / keras-yolo2

Easy training on custom dataset. Various backends (MobileNet and SqueezeNet) supported. A YOLO demo to detect raccoon run entirely in brower is accessible at https://git.io/vF7vI (not on Windows).

MIT License

1.73k stars 784 forks source link

Number of batches remain same after augmentation #424

Open motiraj99 opened 5 years ago

motiraj99 commented 5 years ago

I printed the number of training batches after augmentation using train_batch.len(). But it is same as without augmentation. Why is it so?

rodrigo2019 commented 5 years ago

the batch size is fixed, but your data is augmented