YOLOv5 Torch

This guide explains how to run YOLOv5 with torch backend.

---

(Supports every device and platform except TPU, Full Dynamic Support)

You can use Torch powered detector by specifying the backend parameter.

from cvu.detector import Detector
detector = Detector(classes="coco", backend = "torch"))

Internally, the Detector will load Torchscript (JIT) pretrained Yolov5s weight model.

If you want to run the detector for your custom weights, simply do the following:

• Train Yolov5 on your custom dataset
• Export Weights PyTorch weights to TorchScript

Make sure your model is on the correct device (CUDA can save torchscript in Float16 format which is unavailable/inefficient in many CPUs) while exporting your custom weights. It's recommended to add --half flag for CUDA.

python export.py --weights $PATH_TO_PYTORCH_WEIGHTS --include torchscript

Now simply set parameter weight="path_to_custom_weights.pt in the Detector initialization, and you're ready for inference.