Blur Plates and Faces in Video Files

Blur software processes frames of a video to blur a video file.

Download Scripts

The source code for the utilities is available here, make sure you are in your working directory and download all files.

git clone --depth 1 https://github.com/parkpow/deep-license-plate-recognition.git
cd deep-license-plate-recognition/video-editor
cp env.txt.sample env.txt

Create Environment Variables File

Once you downloaded the video-editor files, you need to set up the environment variables by creating an env.txt file in the same directory as the docker-compose.yml file. You can find a sample env.txt.sample file in the same directory.

Variable	Description	Default
TOKEN	Your API token for Snapshot. Get it from here.	None
LICENSE_KEY	Your Snapshot SDK license key. You can find it here.	None
VIDEO_BLUR	Controls the blur intensity applied. The weakest blur is 10. The strongest is 1.	1
SAMPLE	Controls how many frames are skipped when performing inference. To decrease processing time, Blur performs computationally intensive inference on one frame out of SAMPLE frames seen and then approximates blurring in between. Raise this parameter to speed up processing, lower it to improve quality.	5
LOGGING	The logging level for the API. Available logging levels are DEBUG, INFO, WARNING, ERROR, and CRITICAL	DEBUG

This is how your env.txt file should look like:

env.txt

LOGGING=DEBUG
TOKEN=YOUR_TOKEN
LICENSE_KEY=YOUR_LICENSE_KEY
VIDEO_BLUR=1
SAMPLE=3

note

You can add more configuration parameters at the end of your environment file, using the format PARAMETER=VALUE, for example, REGIONS=in,us

danger

Previous versions of the Blur video utility used the BLUR environment variable to control the blur intensity. This variable is now deprecated. Use VIDEO_BLUR instead.

Build and Run the Docker Image

Once you have set up your environment variables, you can build and run the video-editor image with docker-compose.

docker volume create license # if you do not use Snapshot on this device already
docker-compose up --build

Upon successful build, the API will be available at http://localhost:8081/process-video.

Blur Plates and Faces

To blur plates and faces in a video, pass the video file path and action=blur to the API.

curl -X POST -F "upload=@/user-data/test.mp4" -F "action=blur" http://localhost:8081/process-video

A new video named blur-{camera_id}.mp4 will be created in the output folder.

API Parameters

Parameter	Description	Default
upload	The video file to process	None
action	The action to perform on the video file. options blur, frames and visualization. See the below instructions for the frames and visualization actions.	None

Both parameters are required for the API to work. If any of the parameters are missing, you will get the following response:

{
  "error": "Invalid request."
}

Additional Features

Extract Frames as Images

To extract frames from a video, pass the video file path and action=frames to the API. For example:

curl -X POST -F "upload=@/user-data/test.mp4" -F "action=frames" http://localhost:8081/process-video

The extracted frames will be saved in a folder named {filename}_frames inside the output folder.

Draw Vehicle and Plate Bounding Boxes

This works by uploading every frame in the video to Snapshot.

Visualize plates and vehicles on a video, pass the video file path, and action=visualization to the API. For example:

curl -X POST -F "upload=@/user-data/test.mp4" -F "action=visualization" http://localhost:8081/process-video

A new video named {filename}_visualization.avi will be created in the output folder.