This is a ROS Server served as an Object Detection using Faster_RCNN Method. Given an image and then output bbox information.
This idea comes form FasterRCNN
This code is modified from py-faster-rcnn, but I modified it as a ROS server to do specific task for object Recognition in PROGRESSLAB
To run this demo, you have to install caffe_model as mentioned in py-faster-rcnn in scripts folder and include a pretrained .caffemodel file. For model to download, please contact me [email protected]
- Please recursive clone my repo, make sure you have included caffe-fast-rcnn folder in scripts folder
- In scripts/caffe-fast-rcnn folder, make sure you have changed the branch to faster-rcnn branch, not master branch
- In scripts/caffe-fast-rcnn folder, make a copy of Makefile.config.example as Make.config and uncomment the line: WITH_PYTHON_LAYER := 1
- In scripts/caffe-fast-rcnn folder, add opencv_imgcodecs in file Makefile. It is around the line 174 in LIBRARIES variable. Then run
make
andmake python
. And pip install scikit-image, easydict, protobuf, cython, apt-get install python-opencv - In scripts/data/demo folder, include test images and make sure in client.py in scripts folder has specified the path of these test images.
- In scripts/data/faster_rcnn_models, include pretrained model and make sure in demo.py in scripts folder has specified the path of pretrained model.
- Run demo.py first, then run client.py
The demo video is shown on following youtube link demo
Use --net NET_NAME
to indicate which trained model to use, as being specified in
NETS = {key: (name of folder, model name, pre-trained or not, tuples of class)}