livedetect

README

LiveDetect

LiveDetect is a tool designed to process local video stream captured via camera, and execute machine learning models on every frame using DeepDetect.

Quickstart

Quickstart example for running LiveDetect on a Raspberry Pi 3B+:

• sudo apt install libjpeg-dev

• wget https://github.com/jolibrain/LiveDetect/releases...

• tar xvf livedetect-pi3.tar.gz && chmod +x livedetect

NOTE: If you want to build LiveDetect by yourself, please refeer to the Build section of this README.

First, you need a DeepDetect instance running and accessible from the machine where you want to use LiveDetect.

If you want to run DeepDetect directly on a Raspberry Pi 3, here is a sample command to start a DeepDetect container with only NCNN as back-end (well suited for running directly on a Raspberry Pi).

• docker run -d -p 8080:8080 -v $HOME/models:/opt/models jolibrain/deepdetect_ncnn_pi3

For an AMD64 container, CPU-only, using Caffe:

• docker run -d -p 8080:8080 -v $HOME/models:/opt/models beniz/deepdetect_cpu

Or if you want a GPU build:

• docker run -d -p 8080:8080 -v $HOME/models:/opt/models beniz/deepdetect_gpu

NOTE: for the GPU container, you need to install nvidia-docker.

You're ready for the following examples!

Examples

Face detection + bounding boxes on web preview

./livedetect \
    --port 8080 \
    --host 127.0.0.1 \
    --mllib ncnn \
    --width 300 --height 300 \
    --detection \
    --create --repository /home/pi/face/ \
    --init "https://www.deepdetect.com/models/init/ncnn/squeezenet_ssd_faces_ncnn_300x300.tar.gz" \
    --confidence 0.3 \
    --device-id 0 \
    -v INFO \
    -P "0.0.0.0:8888" \
    --service face \
    --nclasses 2

This command start LiveDetect and tell the DeepDetect instance located at 127.0.0.1:8080 to create a service named face-detect.

This service takes 300x300 pictures for a detection process with the model located at the adress specified by --init, that have 2 classes.

The preview is displayed on port 8888, specified with -P.

V4L uses --device-id as the capture device (here device 0) and verbosity at the INFO level is triggered by -v INFO.

Detection + mask

./LiveDetect \
    --port 4242 --host localhost \
    --width 416 --height 608 \
    --detection \
    --create --repository /home/user/test_mask \
    --mask -e /home/user/test_mask/mask \
    -m "102.9801" -m "115.9465" -m "122.7717" \
    --confidence 0.1 --nclasses 81 \
    --service mask --device-id 1 \
    -P "0.0.0.0:8888" \
    --select-classes -c car -c person -c truck -c bike -c van

This command start LiveDetect and tell the DeepDetect instance located at localhost:4242 to create a service named mask for detection and mask preview.

It uses files under the --repository path with extensions files under the path specified by -e.

Mean values are specified with -m, 81 classes, using the GPU, service should be configured for images of size 416x608 and predictions should be considered only if their confidence is at least 0.1.

Finally, using --select-classes, we specify classes we want to be previewed and processed, using classes names with the -c flags.

Detection + InfluxDB

LiveDetect support InfluxDB, the categories detected and the probability for each category can be pushed to an InfluxDB database, here is the previous example, with the InfluxDB parameters.

./LiveDetect \
    --host 167.389.279.99 --port 4242 \
    --width 300 --height 300 \
    --detection \
    --create --repository /home/user/model/ \
    --confidence 0.1 --nclasses 601 --device-id 1 --gpu \
    --service openimage \
    -v INFO \
    --influx --influx-host http://192.168.90.90:8080 \
    --influx-user admin --influx-pass super-strong-password \
    --influx-db livedetect

To use InfluxDB, you first need to pass the --influx argument, then you can specify the host with --influx-host, the credentials that should be used with --influx-user and --influx-pass, and finally the database name with --influx-db.

Build from source

Requirements

In order to build and use LiveDetect, you need Go.

Create your Go workspace, for example in your home directory:

mkdir $HOME/go

To install Go, download the approriate version for your OS here.

When the download is complete, extract it to install the go toolchain:

tar -C $HOME -xzf go$VERSION.$OS-$ARCH.tar.gz

Add $HOME/go/bin to the PATH environment variable. You can do this by adding this line to your /etc/profile (for a system-wide installation) or $HOME/.profile:

export PATH=$PATH:$HOME/go/bin

Note: changes made to a profile file may not apply until the next time you log into your computer. To apply the changes immediately, just refresh the file with source $HOME/.profile.

Finally, set the GoPATH, for this, refeer to this page.

LiveDetect build

Now that you have the required packages fetch LiveDetect:

go get -u github.com/jolibrain/livedetect

Move in the directory containing the sources:

cd $GOPATH/src/github.com/jolibrain/livedetect

Fetch all dependencies remaining:

go get -u ./...

Finally, build it!

go build .