A YoloV7 Web Service for Person Detection
Older security cameras either don't support AI person detection or use out of date models with lower accuracy. This web service can be used to analyze the snapshots generated during motion events to help determine if they're triggered by people. Additional objects can also be detected, however, only persons will be annotated in the snapshots.
https://github.com/Webchantment/YoloV7WebService/releases
ASP.NET Core Runtime 6.x and .NET Runtime 6.x
https://dotnet.microsoft.com/en-us/download/dotnet/6.0
Use Visual Studio 2022 and compile a Release Build
Run "YoloV7WebService.exe" and visit: http://localhost:5000/swagger
By default the web service runs in CPU mode.
If you want GPU acceleration add the "-gpu" command line option (requires CUDA drivers): "YoloV7WebService.exe -gpu"
Two endpoints are available:
/predictions/{imagePath} - Returns objects detected for a given image in JSON format. If a person is detected, a new annotated image will be created with "-person" appended to the filename. ex) "Image-person.jpg"
/benchmark/{imagePath} - Returns the time in milliseconds it takes to load and run object prediction for a given image.
