Hello.. We tried your training script with portrait images, the training goes fine.. But with pascal VOC, since the masks are different, we uncommented part of the code in loader.py
All these commented lines, we uncommented, as the masks has to be handled differently for Pascal VOC.. Still the output goes bad, even around 22k epochs.. Is this methodology correct or what else needs to be done? And at how many epochs can we expect a decent enough output?
Hello.. We tried your training script with portrait images, the training goes fine.. But with pascal VOC, since the masks are different, we uncommented part of the code in loader.py
All these commented lines, we uncommented, as the masks has to be handled differently for Pascal VOC.. Still the output goes bad, even around 22k epochs.. Is this methodology correct or what else needs to be done? And at how many epochs can we expect a decent enough output?