Model Convert

Model Convert is used to convert YOLOv5 Torchmodel to TensorRT format, which can significantly improve detection performance.

Feature Entry: Toolbox->Machine Learning->Video Analysis->Model Convert.

Training Data Path: Select the folder containing training data generated through Manage Picture Sample.
Models: Select the model to be converted.
Training Picture Size: Set the picture size for model training, which must match the picture size used in model training.
Video Width: Set video width, which must match the width of the video to be detected after conversion.
Video Height: Set video height, which must match the height of the video to be detected after conversion.

Important Notes:

The converted model can only be used to detect videos with the set aspect ratio. For example, if set to 1920*1080, the model can only process videos with resolutions like 1920*1080 or 1280*720.