Day-38 Computer Vision Learning ResNet-38 — Wider or Deeper ResNet? (Image Classification & Semantic Segmentation) by University of Adelaide Follow me for similar post : 🇮🇳 Ashish Patel Interesting Facts : 🔸 It is published in 2019 #JPR, which has already got over 505 citations 🔸 A Good Compromise Between the Depth and Width, Outperforms DeepLabv2, FCN, CRF-RNN, DeconvNet, DilatedNet, Comparable with DeepLabv3, PSPNet. ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://lnkd.in/eFR9Rkp tensorflow : https://bit.ly/3jxJupl pytorch : https://bit.ly/36PBk6q keras :https://bit.ly/2YTRoQs ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 It outperforms the original ResNet in image classification and image segmentation. 🔸 They have used preactivation Resnet with Relu and batch normalization. 🔸 Resolution: To generate score maps at 1/8 resolution, down-sampling operations are removed and dilation rates are increased in some convolutions. 🔸 Max pooling is harmful due to too strong spatial invariance. more in comments #computervision #artificialintelligence #data #datascience
I think better to use ResNet50 or ResNet10, because they are more complex than 38.
🔸 Classifier: One convolution is added to make the channel number equals to number of pixel categories, e.g. 21 for PASCAL VOC 2012, denoted as “1 conv”. 🔸 One more 512-channel convolution can be added at the middle as well, denoted as “2 conv”. #innovation #technology #deeplearning #india For previous post visit this github : https://github.com/ashishpatel26/365-Days-Computer-Vision-Learning-Linkedin-Post
Nice topic 🇮🇳 Ashish Patel! For a holistic overview on CNNs advise my recent article: https://theaisummer.com/cnn-architectures/