Ashish Patel 🇮🇳’s Post

𝗗𝗮𝘆-𝟭𝟬𝟮 Computer Vision Learning 𝗤𝗦𝗔+𝗤𝗡𝗧 — Quantization of Fully Convolutional Networks (Biomedical Image Segmentation) by Univerity of Notre Dame and Huazhong University of Science and Technology Follow me for similar post : 🇮🇳 Ashish Patel Interesting Facts : 🔸 This is a paper in 2018 CVPR with over 50 citations. 🔸 Incremental Quantization on Neural Network, Act as Regularization Term, Reduce Overfitting ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://lnkd.in/ec7mwzJ ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 7-bit quantization on 5 Segmentation FCN (QNT) and 7-bit quantization on 5 Suggestive FCN (QNT) outperforms SA, MultiChannel (Both conference and transaction versions), and CUMedVision2 / DCAN 🔸4.6× and 6.4× reduction on memory for 7-bit and 5-bit quantization respectively. 🔸It is just like a regularization term. #computervision #artificialintelligence #deeplearning