How can I use Pytorch/Tensorflow based custom algorithms for ground truth labeling automation?

Hi! I am trying to automate ground truth labeling in MATLAB. The default algorithms dont work well for my dataset.

Hence, I was looking to use a pytorch based object detection model.

Can someone please guide me to documentation on that?

Answer 1

[Sivylla Paraskevopoulou]

The Image Labeler app enables you to label ground truth data in a collection of images.

If you want to use a PyTorch model, you can co-execute the object detector in Python. For an example on how to how to automate object labeling in the Image Labeler app using an object detection model trained in Python, see Automate Labeling in Image Labeler Using a Pretrained TensorFlow Object Detector.

Alternatively, you can only import TensorFlow and ONNX models into MATLAB by using the importTensorFlowNetwork and importONNXNetwork functions, respectively. You can convert a PyTorch model into the ONNX model format and then, import the ONNX model into MATLAB using the importONNXNetwork function. For an example on how to import a pretrained ONNX object detector, see Import Pretrained ONNX YOLO v2 Object Detector.