Need advice using custom dataset

[ramdhan1989]

Hi this is really good application. I would like to have your advice to use my own dataset. I have images that are very different with widely known dataset such as imagenet containing common object (cars, cat, dog, etc). this is very technical related, it is dynamometer cards that commonly used in oil industry as shown below :
https://knepublishing.com/index.php/KnE-Engineering/article/download/3083/6588/15176
Unlike common object, my dataset always comes in neutral position, there is no rotated, no flipped, no color distortion (just B&W) and always comes in the full part (unlike cars that might exist only part of it in images), etc that might exist during implementation in the real world. So in the common object recognition, I assume such data augmentations are needed to introduce many variations to the model.
Do I need to do same data augmentations used by default in this repo ? or is it ok for BYOL to use no augmentations or minimum augmentations ? what do you think the best solution needed for my own dataset especially in term of augmentation.

thank you

[Shishuii]

Hi, @ramdhan1989 . Can you please share the source of dataset that you are using ?
I am also working on similar problem. It will be very helpful for me , if you can share your dataset of dynacards.