Faster-Rcnn code is run
with the same Faster-Rcnn Faster-Rcnn caffe official also used as a framework.
First cloned into the local item (need to hang Agent)
Make sure to clone with –recursive
--recursive clone git https://github.com/rbgirshick/py-faster-rcnn.git
then enter the lib directory execute make (if it is necessary to switch to make Python2 modify configuration files python3 environment)
cd $ FRCN_ROOT / lib
make
note faster-rcnn can not use a Chinese path, put it in your home directory to properly make
modify the configuration file setup.py CPU version only
Since I am using cpu version, you need to configure makefile.config, then execute the command make -j16
If you need to rely on an error, you need to install and corresponding dependence.
Then execute the command make pycaffe
executed after /data/scripts/fetch_faster_rcnn_models.sh command to download model files, you can also use the browser after the download, unzip the folder on the data
last executed command python2 ./demp.py -cpu
errors need to be modified
../lib/fast_rcnn/nms_wrapper.py:9:#from nms .gpu_nms import gpu_nms
will comment only gpu cpu version
Here reference to someone else's blog https://www.cnblogs.com/justinzhang/p/5386837.html
Detection results are as follows: