Test result transfer

You can label a small number of images and use "Test Result to Annotation" to obtain a large number of pre-labeled images. The specific steps are as follows:

1. Label only a small number of images, train the model, and test the images. (The selected images should be typical, otherwise the training results will be poor.)

2. Check the tested images, use "Test result transfer" for the images that achieve the expected effect, and then click "Save".

3. Retrain the model. The new trained model has a larger sample capacity, so the training effect is generally better.