Detection of rotated objects using TAO

Please provide the following information when requesting support.

• Hardware (T4/V100/Xavier/Nano/etc) GPU-RTX3080
• Network Type (Detectnet_v2/Faster_rcnn/Yolo_v4/LPRnet/Mask_rcnn/Classification/etc) Maskrcnn
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here) Latest TAO version
Does TAO provide training for detection of rotated objects?
For example, like detection of pointer at a guage.
rotated

Any suggestion for detection of rotated object detection using TAO? Something like mmrotate.

For segmentation network, TAO Mask_rcnn can work on such task. For detection networks, currently, TAO has not similar network to do rotated object detection yet. But another toolkit(ODTK) from Nvidia can work on that. See GitHub - NVIDIA/retinanet-examples: Fast and accurate object detection with end-to-end GPU optimization. This repo supports rotated bounding box detections.

Refer to:
https://developer.nvidia.com/blog/detecting-rotated-objects-using-the-odtk/ and
GitHub - NVIDIA/retinanet-examples: Fast and accurate object detection with end-to-end GPU optimization.

More, if the rotate object has only 1 class, currently you can use TAO OCDnet to train to detect the rotate object.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.