Our new paper, "FlashGMM: Fast Gaussian Mixture Entropy Model for Learned Image Compression" is accepted by IEEE VCIP 2025!
This paper accelerates the Gaussian Mixture Model (GMM), an well-known probability model of learned image compression, by ~90x while slighly improving its performance. I have submitted the pre-print version to arXiv. Codes are released under my repo!
I am Shimon Murai, a master course student at Katto Laboratory at Waseda University, Tokyo, Japan. My interests are in Deep Learning, Image Processsing and Neural Image Compression. Check out my papers at Google Scholar. Detailed profiles (in Japanese) are Here!