Aimira Baitieva
Aimira Baitieva
Hello, > I expected tiling to help me use bigger image_sizes with less GPU capacity in order to get more detailed results. I think tiling is mainly supposed to help...
This batch size is specifically mentioned in the paper, and it is necessary to use size 1 for training EfficientAD. For validation, it's also 1 because of how this implementation...
You can find it in the `pretraining.py`.
Can you try it with a lower resolution or fewer images? It can be an out-of-memory error.
Does Padim work if you use it without tiling? It could be just different memory requirements for Padim and PatchCore.
Hello, it should be imagenette dataset, which is about 1.5GB. It is an important part of training EfficientAD which helps model not to overfit on training data.
You can try [custom transforms](https://anomalib.readthedocs.io/en/latest/markdown/guides/how_to/data/transforms.html). Although most anomaly detection papers do not use transforms at all because more robust models can be actually worse at detecting anomalies.
Hello, how big is your dataset and which resolution images are? Both these factors will affect time of training.
Can you try if it works with 256x256? Maybe there is some different problem, especially if the output screen is stuck.
This link in anomalib docs might help you: [how to train with normal images only](https://anomalib.readthedocs.io/en/latest/markdown/guides/how_to/data/custom_data.html#with-only-normal-images), you will have a single folder and synthetic anomalies for validation (or you can skip...