Caffe for PVANET

by Kye-Hyeon Kim, Yeongjae Cheon, Sanghoon Hong, Byungseok Roh, Minje Park (Intel Imaging and Camera Technology)

Introduction

This repository is a fork from BVLC/caffe. Some modifications have been made to run PVANET with Caffe:

• Implemented a new learning rate scheduling based on plateau detection.
• Implemented proposal layer for both CPU and GPU versions.
• Implemented NMS for both CPU and GPU versions.
• Copied RoI pooling layer and smoothed L1 loss layer from py-faster-rcnn
• Applied a hotfix for 'average_loss'