Retina simulation from digital camera source. The following outputs can be generated from one image:
- S,M,L cones output.
- Parvo output by simulating the midget cells outputs
- Magno output by simulating the parasol cells. Temporal aspects are not yet simulated
Parvo and Magno outputs are two different matrices whith values between 0 and 255. The value 0 represent a low pulse frequency (near 0 Hz) and 255 a high pulse frequency (more than 100 Hz).
This project generate a library that use GPU acceleration by using cuda. No pure CPU support is done for now.
If you want more details, check this what I understood page
Requirements:
- docker
- nvidia card with a driver version >= 430 (I did not try below). You do not need cuda to be installed as we compile in a docker
To check if you match the requirement try to run the following command:
sudo docker run --rm --runtime=nvidia nvidia/cuda:10.0-devel nvcc --version
This project can be build in a docker environment. To start it run
./scripts/start_dev.sh
mkdir build
cd build
cmake ..
make
An example is provided
./gpuretina_viz -i ../videos/<some_video>.mp4