David Charatan

I haven't seen this happen before. What happens when you enable fault handler? ``` import faulthandler; faulthandler.enable() ``` Also, what environment are you running on?

You'll have to add `import faulthandler; faulthandler.enable()` at the very top of `flowmap/overfit.py` so that it gets enabled when running the code. If you do this, more information about the...

This seems to be a problem with the CoTracker code from Torch Hub, but I'm not sure why it's segfaulting. Here are a few things to try: - Using `requirements_exact.txt`...

You can try appending `+experiment=low_memory` to the command you use to run FlowMap. However, FlowMap isn't designed to run on video sequences of many hundreds or thousands of frames. There...

See #4 and #13 for a few more comments on memory usage!

@phongnhhn92 Be careful—we found that CoTracker v2 produces less accurate tracks than v1. This might impact performance more than you would like.

Yes, to reproduce the results, you'll need CoTracker V1's weights. It might be worth trying out the steps [here](https://github.com/dcharatan/flowmap/issues/23#issuecomment-2123097547) to avoid having to manually download the weights.

If the `image_shape` is just one number, it's the approximate total number of pixels (so width times height).

Yes, FlowMap will resize the input images until they're roughly at the target resolution. There are two resolutions that FlowMap uses: 1. The resolution at which the optimization runs, which...

Is this an issue with the code at head? It seems the number of workers is hard-coded to 0: https://github.com/dcharatan/flowmap/blob/2be1b9ef8a22513da99b1d7611362f33f9f6481a/flowmap/dataset/data_module_overfit.py#L29