How to merge tiled inference results?

[ok4d4]

Question

Hello, how do you merge the inference results in the NMS method?

[github-actions[bot]]

Hello @ok4d4, thank you for your interest in our work! Ultralytics has publicly released YOLOv5 at https://github.com/ultralytics/yolov5, featuring faster, lighter and more accurate object detection. YOLOv5 is recommended for all new projects.

If this is a bug report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom model or data training question, please note that Ultralytics does not provide free personal support. As a leader in vision ML and AI, we do offer professional consulting, from simple expert advice up to delivery of fully customized, end-to-end production solutions for our clients, such as:

• Cloud-based AI systems operating on hundreds of HD video streams in realtime.
• Edge AI integrated into custom iOS and Android apps for realtime 30 FPS video inference.
• Custom data training, hyperparameter evolution, and model exportation to any destination.

For more information please visit https://www.ultralytics.com.

[glenn-jocher]

@ok4d4 small objects detected twice in overlapping regions of tiles are automatically merged in NMS. There is nothing special required, simply normal NMS.

[ok4d4]

Isn't there a possibility that objects may not be detected if they are on the boundary of the grid?

[glenn-jocher]

@ok4d4 there's a possibility that objects may not be detected at all locations in the image.

[ok4d4]

Of course you're right. When cropping an image, should I overlap parts of it?

[glenn-jocher]

@ok4d4 yes that's what I mean, if you overlap for example 10% each direction on a 640 crop, then all objects smaller than 64 pixels across should be detected at least once without any breaks. Larger objects may be cut in pieces, but then again the standard use case for sliding window inference is for detecting small objects in large images, so large objects should be less of an issue.

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.