<a href="https://clay-foundation.github.io/model/specification.html" rel="nofollow">https://clay-foundation.github.io/model/specification.html</a><p>> Clay v0 is a self-supervised modified vision transfer model trained on stacks of Sentinel-2, Sentinel-1 & DEM data. It is trained as a Masked Autoencoder (MAE) to reconstruct the original image from a masked image.<p>> Each data entry is a stack of 10 bands of Sentinel-2, 2 bands of Sentinel-1 & 1 band of DEM data. The model is trained with 3 timesteps of data for each location, with a total of 1203 MGRS tiles globally distributed, each of size 10km x 10km. The data was collected from the Microsoft Planetary Computer.<p>> The model was trained on AWS on 4 NVIDIA A10G GPUs for 25 epochs (~14h per epoch) in December 2023.<p>Also useful: <a href="https://clay-foundation.github.io/model/model_embeddings.html#format-of-the-embeddings-file" rel="nofollow">https://clay-foundation.github.io/model/model_embeddings.htm...</a><p>> The embeddings file utilizes the following naming convention:<p>> {MGRS:5}_{MINDATE:8}_{MAXDATE:8}_v{VERSION:3}.gpq<p>> Example: 27WXN_20200101_20231231_v001.gpq<p>MGRS is Military Grid Reference System. I believe 5 characters corresponds to about 60 miles by 60 miles.<p>So presumably the result is an embedding vector representing e.g. a specific spot on earth around a specific 3-year period, such that you can compare that embedding vector with other points to get an idea for what's similar and maybe categorize them against known vectors for things like had-a-lot-of-deforestation?