search

chenyuntc / simple-faster-rcnn-pytorch

A simplified implemention of Faster R-CNN that replicate performance from origin paper
Other
3.97k stars 1.14k forks source link

How to train Faster R-CNN on my own custom dataset and changing the RPN loss functions? #241

Open meshghi opened 3 years ago

meshghi commented 3 years ago

Hello. I want to:

1- train Faster RCNN on my own custom dataset. 2- use the pre-trained Faster RCNN on the VOC2007 as the initial weights to train it then on my own custom dataset. 3- modify the RPN network with different loss functions that I have come up with.

I will greatly appreciate it if you could help me/let me know how to achieve the above goals.

meshghi commented 3 years ago

Hello again,

I changed most of the code to work with my custom dataset. But, I'm stuck with handling the "difficult" trining objects. My custom dataset doesn't have any "difficult" object and thus I get errors during training in areas like below:

difficult.append(int(obj.find('difficult').text))

I was wondering how I could fix that? Thanks

WuXiaoyu55 commented 2 years ago

Hello, I encountered the same problem, did you solve it？ difficult.append(int(obj.find('difficult').text))