Ashish Patel 🇮🇳’s Post

Day-31 Computer Vision Learning 𝗣𝗿𝗲-𝗔𝗰𝘁𝗶𝘃𝗮𝘁𝗶𝗼𝗻 𝗥𝗲𝘀𝗡𝗲𝘁 with Identity Mapping with 𝟭𝟬𝟬𝟬 𝗟𝗮𝘆𝗲𝗿𝘀 (Image Classification) by Microsoft Follow me for similar post : 🇮🇳 🇮🇳 Ashish Patel 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗙𝗮𝗰𝘁𝘀 : 🔸 It is published in 2016 #ECCV, which has already got over 4972 citations 🔸 Resnet goes from 101 to 1202 Layes. 🔸 Error rate degradation from 6.43% to 7.93% 🔸 A chained residual pooling is also introduced which captures rich background context in an efficient manner. ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://lnkd.in/etCDUBf lua : https://bit.ly/3tdkdFd Pytorch : https://bit.ly/3tbpb5n Tensorflow and Keras : https://bit.ly/3cqSlb1 ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 Resnet 1001 Containes Direct connection, scale connect , gating, 1x1 convolution 🔸 the optimization is further eased because f is an identity mapping 🔸 using BN as pre-activation improves regularization of the model. #artificialintelligence #computervision #innovation #india