Jacob Prince-Bieker

Optical flow baseline is running now for the 12 sat channels, where the flow is determined for each channel separately. Preliminary results is that the MSE is ~0.9 for optical...

Plots from the running job. The MSE depends a lot on the specific channel for optical flow. There also seems to be large jumps for certain channels after an hour,...

The timedelta between all the future timesteps is 5min, like it should be, so not sure for the jump after timestep ~7, but it doesn't seem to be an issue...

Things to try: Take the mean of a few pairs of previous images to do the optical flow, and get the optical flow between the last frame and the previous,...

Yeah, I am too. Computing the Optical flow as the average of the optical flow of the last 5 pairs of images gives more of a difference for the MSE....

While actually using DARDAR doesn't seem very feasible for making a model, this product: https://navigator.eumetsat.int/product/EO:EUM:DAT:MSG:OCA was calibrated off DARDAR and gives more detailed cloud properties, but is only every hour....

Also could be useful are these products: https://data.eumetsat.int/product/EO:EUM:DAT:METOP:IASIL1C-ALL?query=IASI%20L1C&s=extended and https://data.eumetsat.int/product/EO:EUM:DAT:METOP:AVHRRL1?query=&s=simple that give more detailed cloud and data streams. So in training we could give higher resolution masks, or more cloud...

Also, these products have the advantage of being from the product store, so we can download them without needing to wait some indeterminate time, like the other product.

Try it with https://github.com/facebookresearch/xcit which is available from `timm`

The PR showing the errors from the linter for SatFlow is here: #96