Quantcast
Channel: TAO Toolkit - NVIDIA Developer Forums
Viewing all articles
Browse latest Browse all 409

Certain classes in my object detection model are being mislabeled

$
0
0

Please provide the following information when requesting support.

• Hardware (T4/V100/Xavier/Nano/etc) Geforce rtx 3090
• Network Type Yolo_v4 (Detectnet_v2/Faster_rcnn/Yolo_v4/LPRnet/Mask_rcnn/Classification/etc)
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here) command not recognized
• Training spec file(If have, please share here) yolo_v4_train_resnet18_kitti.txt
• How to reproduce the issue ? (This is for errors. Please share the command line and the detailed log here.) NA

My object detection model, trained in yolo_v4 is showing a high detection rate. However, there are 3-4 object (classes) that are consistently mislabeled with the same label. I don’t think it is an incorrect detection but rather a mislabeling based on the consistency of the objects affected.

It seems that the class being mislabeled is always 2 classes apart from the class it is labeled for in the class mapping list. I should also mention that there are 120 classes which could possibly play into the equation, and that at this point I am training with only about 550 images.

I have tried a number of possible solutions with no success and I’m running out of ideas. Can you suggest some options I might check for possible solutions or questions I should consider?

Wayne

2 posts - 2 participants

Read full topic


Viewing all articles
Browse latest Browse all 409

Trending Articles