0

While training my own custom dataset with DeepMask model, accuracy is reaching 99.5 percent. When I compute proposals using this model, the results are good but still not commensurate with the accuracy.

I am doing wound detection and I have 1500 images. Maybe it is because my custom dataset has only one class.

So if I want to train my custom dataset with only one class, what approach should I follow? Or should I solely rely on IoU and Recall matrix?

IOU: mean 065.19 median 066.51 suc@.5 087.25 suc@.7 038.86 | acc 097.31 | bestmodel *
abubakr
  • 15
  • 3
user195812
  • 1

1 Answers1

0

Ok I got the answer from this link.

I was doing:

First data augmentation on the data, then splitting the data into training and validation set.

The answer is: "Never do this thing, as you will get leakage. for example assume the augmentation is a 1-pixel shift left. if the split in not augmentation aware, you may get very similar data samples in both train and validation."

So I need to do this:

"First split the data into training and validation sets, then do data augmentation on the training set. You use your validation set to try to estimate how your method works on real world data, thus it should only contain real world data. Adding augmented data will not improve the accuracy of the validation. It will at best say something about how well your method responds to the data augmentation, and at worst ruin the validation results and interpretability."

abubakr
  • 15
  • 3