TAO classification pyt - Very Low accuracy on custom dataset

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

Please provide the following information when requesting support.

• Hardware (T4/V100/Xavier/Nano/etc) : A6000
• Network Type (Detectnet_v2/Faster_rcnn/Yolo_v4/LPRnet/Mask_rcnn/Classification/etc) : Classification
• TLT Version (Please run “tlt info --verbose” and share “docker_tag” here) : 5.3.0
• 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 m trying an custom binary image classification on the image classification pyt notebook, modified the spec’s as per that but the accuracy of the model isnt improving, What could be reason ? Attaching the spec’s and training status json file.

spec.txt (1.1 KB)
status.txt (43.1 KB)

2

Please try

  1. check your images’ resolution and tune scale: 224.
  2. use AdamW optimizer.
  3. more augmentation methods
  4. tune learning rate.

More example training spec files can be found in tao_pytorch_backend/nvidia_tao_pytorch/cv/classification/experiment_specs at main · NVIDIA/tao_pytorch_backend · GitHub.

3

There is no update from you for a period, assuming this is not an issue anymore. Hence we are closing this topic. If need further support, please open a new one. Thanks

4

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