How to train faster RCNN with own region proposals with own dataset in caffe?

[rafayzia]

Hi there. I am new to this forum, also i am new to deep learning. I was successful in running faster RCNN on pascal VOC dataset as described in the code section. If i am not wrong Faster RCNN generates region proposals using selective search or other algorithms. But in my problem, i have my own region proposals. So i want to bypass the RPN and put my region proposals in that place after which ROI pooling takes place. My question is that to do above mentioned tasks, which files i need to change and where to change? Secondly if i have 4 proposals for an image then how should I prepare data in caffe?

[warsi22]

https://www.youtube.com/watch?v=Rgpfk6eYxJA

[alice-cool]

I also want to have_rpn is false, do you solve the remand? If you have methods, please help me.