fenrir

Basic Event Visualization Demo

SSH onto fenrir with X forwarding enabled

$ ssh -X ncs@fenrir
$ cd Eben/fenrir

There should exist a python virtual environment, activate with the following command:

ncs@fenrir:~/Eben/fenrir$ source ~/ebenvenv/bin/activate

Then to see events:

python scripts/dv_processing_example.py --calib_json fenrir_calibration.json --hot_pixel_dir hot_pixels_fenrir

Pass the flag --no_filters to see unfiltered events.

Depth Perception Demo

From the project directory, simply run the executable:

ncs@fenrir:~/Eben/fenrir$ build/SlamDemo --calibration-json fenrir_calibration.json --hot-pixels-dir hot_pixels_fenrir --sbm-num-threads 12  --max-events-in-buffer 24000

--sbm-num-threads is one of many adjustable flags, in this case controlling the number of threads dispatched for each stereo block matching operation.

If the executable does not exist for some reason, rebuild it as follows.

Depth Perception Demo Rendered on Client

Rendering on fenrir and sending the GUI over SSH creates unnecessary overhead. Run the client script from the directory of this repository

python scripts/fenrir_vis_client.py --calib_json fenrir_calibration.json

The dependencies for the above script are numpy, opencv-python, and zmq. The calibration file is so that the depth color key can be calculated correctly. The script can write a video if desired:

python scripts/fenrir_vis_client.py --calib_json fenrir_calibration.json --record_video --video_name <DEFAULT=recording.mp4>

The video will be saved after you interrupt the application with Ctrl+C. Run the demo on the robot but now pointing it towards the local IP address of the client; if running from hugin:

ncs@fenrir:~/Eben/fenrir$ build/SlamDemo --calibration-json fenrir_calibration.json --hot-pixels-dir hot_pixels_fenrir --sbm-num-threads 12  --max-events-in-buffer 24000 --client-addr 172.16.222.30

Starting the demo on the robot before opening the client listener will cause depth frames to backlog on the robot, which will look really weird when you run the visualization.

Build Instructions

Assuming gcc and cmake are installed and working and that dependencies from the next secion have been installed (which is likely unless the system was wiped),

Make sure the build directory exists and enter it

ncs@fenrir:~/Eben/fenrir$ mkdir build
ncs@fenrir:~/Eben/fenrir$ cd build

Configure cmake to build an optimized executable and run the compilation:

ncs@fenrir:~/Eben/fenrir/build$ cmake -DCMAKE_BUILD_TYPE=Release ../
ncs@fenrir:~/Eben/fenrir/build$ make

Now the executable SlamDemo should be present in the build directory.

Dependencies

The following should already be installed on the system, but if something breaks, re-installation may be necessary.

1. cnpy

2. OpenCV

3. dv-processing

4. Boost (sudo apt install libboost-all-dev)

5. ZeroMQ (sudo apt install libzmq3-dev cppzmq-dev)