ibm-nasa-geospatial

Prithvi-EO-2.0 is the second generation EO foundation model jointly developed by IBM, NASA, and Jülich Supercomputing Centre. Prithvi-EO-2.0 is based on the ViT architecture, pretrained using a masked autoencoder (MAE) approach, with two major modifications as shown in the figure below. First, we replaced the 2D patch embeddings and 2D positional embeddings with 3D versions to support inputs with spatiotemporal characteristics, i.e., a sequence of T images of size (H, W). Our 3D patch embeddings consist of a 3D convolutional layer, dividing the 3D input into non-overlapping cubes of size (t, h, w) for time, height, and width dimensions, respectively. For the 3D positional encodings, we first generate 1D sin/cos encodings individually for each dimension and then combine them together into a single, 3D positional encoding. Second, we considered geolocation (center latitude and longitude) and date of acquisition (year and day-of-year ranging 1-365) in the pretraining of the TL model versions. Both encoder and decoder receive time and location information for each sample and encodes them independently using 2D sin/cos encoding. They are added to the embedded tokens via a weighted sum with learned weights: one for time and one for location and separate weights for encoder and decoder. Since this metadata is often not available, we added a drop mechanism during pretraining that randomly drops the geolocation and/or the temporal data to help the model learn how to handle the absence of this information. | Model | Details | Weights | |------------------------|-----------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------| | Prithvi-EO-2.0-tiny-TL | Pretrained 5M parameter model with temporal and location embeddings | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-tiny-TL | | Prithvi-EO-2.0-100M-TL | Pretrained 100M parameter model with temporal and location embeddings | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-100M-TL | | Prithvi-EO-2.0-300M | Pretrained 300M parameter model | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-300M | | Prithvi-EO-2.0-300M-TL | Pretrained 300M parameter model with temporal and location embeddings | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-300M-TL | | Prithvi-EO-2.0-600M | Pretrained 600M parameter model | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-600M | | | Prithvi-EO-2.0-600M-TL | Pretrained 600M parameter model with temporal and location embeddings | https://huggingface.co/ibm-nasa-geospatial/Prithvi-EO-2.0-600M-TL | The models were pre-trained at the Jülich Supercomputing Centre with NASA's HLS V2 product (30m granularity) using 4.2M samples with six bands in the following order: Blue, Green, Red, Narrow NIR, SWIR, SWIR 2. Benchmarking We validated the Prithvi-EO-2.0 models through extensive experiments using GEO-bench. While Prithvi-EO-2.0-100M-TL performs lower than it's bigger counterparts (model params visualized by point size), it needs less compute indicated by the GFLOPs. Demo and inference We provide a demo running Prithvi-EO-2.0-300M-TL here. There is also an inference script (`inference.py`) that allows to run the image reconstruction on a set of HLS images assumed to be from the same location at different timestamps (see example below). These should be provided in chronological order in geotiff format, including the channels described above (Blue, Green, Red, Narrow NIR, SWIR 1, SWIR 2) in reflectance units. You can finetune the model using TerraTorch (`terratorch>=1.1` required). Examples of configs and notebooks are provided in the project repository: github.com/NASA-IMPACT/Prithvi-EO-2.0. Example Notebooks: Multitemporal Crop Segmentation >>Try it on Colab (Choose T4 GPU runtime) Landslide Segmentation >>Try it on Colab (Choose T4 GPU runtime) Carbon Flux Prediction (Regression) If you plan to use Prithvi in your custom PyTorch pipeline, you can build the backbone with: Find more information on model usage in our Prithvi Docs. Your feedback is invaluable to us. If you have any feedback about the model, please feel free to share it with us. You can do this by starting a discussion in this HF repository or submitting an issue to TerraTorch on GitHub. If this model helped your research, please cite Prithvi-EO-2.0 in your publications.