Yolov4 low map

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware (T4/V100/Xavier/Nano/etc) T4•
• Network Type (Detectnet_v2/Faster_rcnn/Yolo_v4/LPRnet/Mask_rcnn/Classification/etc) Yolo_v4•
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here) tao toolkit
• Training spec file(If have, please share here)
• How to reproduce the issue ? (This is for errors. Please share the command line and the detailed log here.)

I used cspdarknet53 and darknet53(from ngc catalog) as backbone, but I got low map. How to convert pretrained darknet(alexeyab’s backbone) to use it in toa toolkit? Is it possible to use trained model(using Alexeyab github), in tao toolkit? How to convert .weights file to hdf5 file (or tlt file)?

2

No, it is not compatible with TAO.
For low mAP, it is not expected. Please check if you use kmeans to generate anchor shapes against your training dataset.

3

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