Summary This repository compiles research papers on remote sensing presented at CVPR 2024, serving as a centralized resource for the latest advancements. It targets researchers, engineers, and practitioners, offering direct access to state-of-the-art methodologies, code, and datasets to accelerate development in areas like land cover mapping, object detection, and Earth observation foundation models.

How It Works The collection highlights diverse methodologies in remote sensing research. Key themes include large-scale foundation models (e.g., S2MAE, SkySense) using spatial-spectral pretraining and transformers. Diffusion models are prevalent for image synthesis, super-resolution, and nowcasting. Vision-language models (GeoChat) and multi-modal learning are increasingly used for interpretation and forecasting. Other techniques encompass advanced CNNs, self-supervised learning, and specialized architectures for BEV fusion and oriented object detection.

Highlighted Details

• S2MAE: A spatial-spectral pretraining foundation model for spectral remote sensing data (Oral paper).
• GeoChat: A grounded large vision-language model for remote sensing interpretation.
• SkySense: A multi-modal foundation model for universal interpretation of Earth observation imagery.
• Sat2Scene: Generates 3D urban scenes from satellite images using diffusion models.
• WildlifeMapper: Detects and identifies multiple species from aerial imagery.
• Radar Fields: Extends radiance fields to SAR data, a novel application for SAR imagery.

Maintenance & Community This repository is a community-curated list of academic publications from CVPR 2024 and its workshops, reflecting current research in computer vision and remote sensing. No specific maintenance details or community channels are provided.

Licensing & Compatibility The repository itself lacks a specified license. Users must refer to the individual licenses of linked papers, code repositories, and datasets, which may vary. Commercial use or closed-source linking depends on these individual component licenses.

Limitations & Caveats This is a curated list of research papers, not an integrated software project. Users must independently evaluate, download, and integrate individual solutions. Content is specific to CVPR 2024 research, representing a snapshot of the field at that time.