Hi, is there a face detection plugin/model available for deepstream sdk 4?
There is a pre-retrained detection model for face in jetson_inference GitHub:
[url]jetson-inference/detectnet-console-2.md at master · dusty-nv/jetson-inference · GitHub
You can change the primary-gie model path and update the bbox parser with the following:
Yes there is, but aren’t the models trained using pytorch? where does TensorRT come in play here, which is actually a requirement for running any inference on a stream.
The model is trained with Caffe and it is supported directly by the Deepstream.
You can just update the .caffemodel and .prototxt path, Deepstream will convert it into TensorRT engine for you.
I am using DeepStream 4.0 and I followed example of nvdsinfer_customparser. I set:
But, I’m getting “Failed to parse bboxes using custom parser function” error.
Are you going to update the detectnet example to work with DeepStream 4.0?
I’m going to check this parser.
Will update more information with you later.
Any update? I also got the same error.
Please open a new topic for DeepStream SDK 5.0. Thanks