Accurate prediction box custom vision

Hi, thanks for clarifying. The following document describes how to improve custom vision model. Based on the information you provided above, my recommendation would be to:

• Be aware of input image requirements
• Provide at least 30 images per tag
• Add more images and balance data; retrain
• Add images with varying background, lighting, object size, camera angle, and style; retrain
• Use new image(s) to test prediction
• Modify existing training data according to prediction results (you can use prediction images for further training)

Hope this helps!

Thanks for answering. I already have tried most of the points you mentioned but I also see some things I haven't tried. I appreciate your help!