using custom dataset

[skpawar1305]

📚 Documentation Improvement

import torch

class MyCustomDataset(torch.utils.data.Dataset):
    def __init__(self, train: bool, image_size: int):
        ...

    def __getitem__(self, item):
        ...
        return inputs, targets # Or inputs, targets, additional_batch_items

Hello, the above class is from the documentation to create a Custom Dataset without recipe. Without any example, it is very hard to understand what should I write, for a custom pose dataset. Anyway, I converted yolonas_pose_n recipe to code with the given script, which is a big help to be honest, and then simply modified data_loaders. Can you please write the above example class with a better example?

Thank you

I'll attach the modified recipe to code just in case train_coco2017_yolo_nas_n.zip

[BloodAxe]

Thanks for reporting!

For pose estimation you can refer to this Notebook on how to make a new dataset: https://github.com/Deci-AI/super-gradients/pull/1831