cv

Image detection tools for security cameras

At the moment, this has a very rudimentary python script that leverages Darknet Yolo and opencv to process one or more videos and figure out if any objects are detected. This is pacaged to run as a Docker Container to make it a little easier to get all the bits and pieces built on various systems.

Example Video

Here's an example video that was detected by this utility from the driveway camera at our cabin in Lake Tahoe

Example usage

docker build -t dhiltgen/cv:cuda .

INST=1
PROC=gpu
docker stop camera${INST}-${PROC}
docker rm camera${INST}-${PROC}
docker run -d \
	--name camera${INST}-${PROC} \
	-v /videos/motioneye:/videos/motioneye \
	--cpus 0.25 \
	--restart always \
	$(if [ ${PROC} = "gpu" ]; then echo "--gpus all"; fi) \
	dhiltgen/cv:${PROC} \
	/videos/motioneye/

docker run --rm -it \
	-v /videos/motioneye:/videos/motioneye \
	--entrypoint gen_thumbs \
	dhiltgen/cv:cpu \
	/videos/motioneye/