Deci-AI / super-gradients

Easily train or fine-tune SOTA computer vision models with one open source training library. The home of Yolo-NAS.
https://www.supergradients.com
Apache License 2.0
4.54k stars 496 forks source link

Auto Augmentation Problem #1656

Closed 7Vyshak7 closed 10 months ago

7Vyshak7 commented 10 months ago

💡 Your Question

I have trained a YOLO-NAS L model for some Pavement markings like Crossmarks, Left arrow, Right arrow, Bike lane, Junction etc. While testing the model on Unseen data the right and left arrow is getting confused. Most right arrow is been identified as Left and vice versa. So if there is some internal Flipping or other augmentation are there, where can I find that, and I want to remove all auto augmentations

Versions

No response

BloodAxe commented 10 months ago

You probably want to remove this augmentation https://github.com/Deci-AI/super-gradients/blob/master/src/super_gradients/recipes/dataset_params/coco_detection_yolo_nas_dataset_params.yaml#L28

7Vyshak7 commented 10 months ago

@BloodAxe Thank You, This solves my issue...

BloodAxe commented 10 months ago

Glad it helped)

tdengs commented 9 months ago

Hi @BloodAxe , is there a way to remove it in the code? i.e. I tried the following but it's not working

image