Open wzleong opened 8 years ago
Hi,@wzleong you can try this version(https://github.com/owphoo/caffe_fast_rcnn) which supports ssd, fast-rcnn and faster-rcnn. Good luck!
Downgrade to CUDNN 4.0, it did the trick for me
I forked the caffe-fast-rcnn-repo and updated the faster-rcnn branch to support cudnn 5.0 and ubuntu 16.04: https://github.com/nils489/caffe-fast-rcnn Be sure to checkout the faster-rcnn branch
Hi, would like to know is CUDA8.0 and CUDNN5.0 supported for FasterRCNN. I have tested that it works for the latest Caffe model. But when I 'make all' for the py-faster-rcnn, there are errors flagged during GPU compilation (mostly on the cuDNN related files for example: caffe-fast-rcnn/src/caffe/layers/cudnn_conv_layer.cu, caffe-fast-rcnn/src/caffe/layers/cudnn_tanh_layer.cu, etc
Thanks. Would like to know is there any workarounds for it.