neural_cam_ros is basically an extension of neural-cam except wrapped with ROS. The backend framework used is darknet developed by Joseph Redmon. You can visit his site here: https://pjreddie.com/darknet/yolo/. This module is designed to quickly test the state-of-the-art obstacle detection for moving robots. No training option is available in this module.

• basic ros-wrapper for obstacle detection is complete
• includes link for pre-trained weights/configuration files

• requires pre-trained weights, cfg files
• rename the parent directory to neural_cam_ros
• open another terminal and type "rostopic echo /obstacle" to see detected objects been published
• edit/launch multiple cameras using multicam.launch file
• i had pre-trained a set of weights to detect road obstacles, there are three classes: 1 - person, 2 - bike and 3 - vehicles. you can download them from this link: https://drive.google.com/open?id=0Bzdkh7sucheCb1V1bG5OWWZZTGs. Do take note that is set of weights is trained with cuDNN v5, be sure to download that from nvidia website. Otherwise, there will be no detections!

• cmake 2.8 or above
• ROS indigo or above
• runs on opencv 3.1 (optional if you only intend to poccess images)
• ubuntu 14.04 above
• cuda 7.5 above (modify the CMakeList.txt if you dont have a GPU!)
• cuDNN v5 (download from nvidia official website, need cuda 8.0)
• latest nvidia graphic driver 367.48

requires ROS to run rosmake

  $ roscd neural_cam_ros
  $ rosmake
  $ roslaunch neural_cam_ros multicam.launch

if cmake display of not able to find "curand/cublas/cudnn", be sure to add the following symbolic link:

   sudo ln -s /usr/local/cuda/lib64/libcurand.so /usr/lib/libcurand.so
   sudo ln -s /usr/local/cuda/lib64/libcublas.so /usr/lib/libcublas.so
   sudo ln -s /usr/local/cuda/lib64/libcudnn.so /usr/lib/libcudnn.so