Fraunhofer HHI launches a 𝗻𝗲𝘄 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗴𝗿𝗼𝘂𝗽 on 𝙇𝙚𝙖𝙧𝙣𝙚𝙙 𝘿𝙖𝙩𝙖 𝘾𝙤𝙢𝙥𝙧𝙚𝙨𝙨𝙞𝙤𝙣!

Led by Prof. Jona Ballé, the team explores fully data-driven approaches to image and video compression, combining neural networks with models of human visual perception.

Jona’s work on 𝗡𝗼𝗻𝗹𝗶𝗻𝗲𝗮𝗿 𝗧𝗿𝗮𝗻𝘀𝗳𝗼𝗿𝗺 𝗖𝗼𝗱𝗶𝗻𝗴 played a key role in the recently finalized 𝗝𝗣𝗘𝗚 𝗔𝗜 𝘀𝘁𝗮𝗻𝗱𝗮𝗿𝗱.

🔗 More about the new research group: https://www.hhi.fraunhofer.de/en/press/news/new-research-group-learned-data-compression-led-by-dr-jona-balle-at-fraunhofer-hhi.html

#NeuralCompression #LearnedCompression #JPEGAI

📑 𝗡𝗘𝗪 𝗣𝗨𝗕𝗟𝗜𝗖𝗔𝗧𝗜𝗢𝗡❗

Our paper explores how #NeuralCompression can transform #EarthObservation by reducing data size while keeping critical insights & will be published in IEEE GRSM in June 2025!

Read it here: https://embed2scale.eu/download/lossy-neural-compression-for-geospatial-analytics-a-review/?wpdmdl=1573&masterkey=LCuRUVlH-A5iKh-3mHF4lK2dD9MOhdKEZMPZF3pVbV4rhnn4yu6yVi7l9WKmhXoWMz2Tu-KwVesCoYVfmJp6lvKFpwoaKekY4nbQdOjcbEo

✅ AI-driven data compression
✅ Self-supervised learning & foundation models
✅ Real-world applications in EO & climate modeling

This is a step forward in making EO data workflows more efficient & scalable!

@[email protected]

AMD To Introduce Neural Texture Compression Technology Next Week #ai #airobotics #amd #neuralcompression

https://www.lowyat.net/2024/325466/amd-to-introduce-neural-texture-compression-technology-next-week/

📢 📢 New Feature in #NeuralCompression repo: Bits-Back compression for diffusion models!
Compress image data 🖼️ using diffusion models at an effective rate close to the (negative) ELBO.

See: https://github.com/facebookresearch/NeuralCompression/tree/main/projects/bits_back_diffusion

Some context ⏩ [1/3]