Hello Dr,
i tried to contact you on linkedin but seems like you aren't there.
First of all thanks for such a great repo that surely helps alot.
BTW i am doing semantic segmentation on a small dataset of endoscopic images. i am able to train a U-NET using segmentation models. i have an image like this:
and respective mask for 5-classes excluding background:
i have increased the number of training images and masks using augmentation but still not getting good results.
i have training iou of 0.53 while testing iou is just 0.45. i am using sigmoid activation function as i needed multi-output ( there is overlap in masks)
please tell me how to improve my IOU in this case.
can i use same augmentation (that i used to increase dataset) during training to overcome overfitting.
Hello Dr, i tried to contact you on linkedin but seems like you aren't there. First of all thanks for such a great repo that surely helps alot. BTW i am doing semantic segmentation on a small dataset of endoscopic images. i am able to train a U-NET using segmentation models. i have an image like this:
and respective mask for 5-classes excluding background:
i have increased the number of training images and masks using augmentation but still not getting good results. i have training iou of 0.53 while testing iou is just 0.45. i am using sigmoid activation function as i needed multi-output ( there is overlap in masks)
waiting for your kind response Thanks :)