Validation loss

[viethoang303]

Hello authors and everyone.

I trained HSnet on my own dataset about echocardigraphy images, but validation loss fluctuates a lot and model converge early. I trained model with learning rate 1e-3 and I looked Fig 9 in paper about this model, validation loss also take 10-30 epochs to achieve higher validation IoU

So, what problem I got. Please help me.

Thank

[juhongm999]

Perhaps the size of your dataset is too small, making the model converge quickly. How many images are there in your dataset?