Ashish Patel 🇮🇳’s Post

Day-39 Computer Vision Learning SqueezeNext — Hardware-Aware Neural Network by University of California, Berkeley Follow me for similar post : 🇮🇳 Ashish Patel Interesting Facts : 🔸 It is published in 2018 #CVPRW, which has already got over 114 citations 🔸 It outperforms AlexNet, VGGNet, SqueezeNet, MobileNetV1 With Lower Complexity or Less inference Time 🔸 It achieves VGG-19 accuracy with only 4.4 Million parameters, 31× smaller than VGG-19. ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://lnkd.in/e2GXbit tensorflow: https://bit.ly/3pZsM4x pytorch: https://bit.ly/2YYDSLa keras: https://bit.ly/3rsbKfG ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 SqueezeNext contains a total of 23 layers (including the fully connected layer), and the main network structure is divided into two structures, SqNxt-23 and SqNxt-23v5. 🔸 The difference between 1.0 and 2.0 is that the number of channels is doubled. 🔸 The difference between traditional and v5 is the difference in network structure Depth. 🔸 With G means depth-wise conv is used, and without it means traditional convolution. #computervision #artificialintelligence #analytics