Day-29 Computer Vision Learning Dilated Residual Network (DRN) (Image Classification and Semantic Segmentation) by Princeton University and Intel lab Follow me for similar post : 🇮🇳 Ashish Patel Interesting Facts : 🔸 It is published in 2017 #CVPR, which has already got over 741 citations 🔸 After Publishing Dilated Net in #ICML 2016 Author has invented new method which can help both Image Classification and Segmentation. ------------------------------------------------------------------- 𝗔𝗺𝗮𝘇𝗶𝗻𝗴 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 : https://lnkd.in/eUpNgbJ Pytorch : https://bit.ly/3aiBpAR Tensorflow : https://bit.ly/2MeiFua Keras : https://bit.ly/2NLwUqO ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 A naive approach is to simply remove subsampling (striding) steps in the network in order to increase the resolution of feature map. However, this also reduces the receptive field which severely reduces the amount of context. such reduction in receptive field is an unacceptable price to pay for higher resolution. more in comments #artificialintelligence #computervision #technology #india
For all previous post follow this link : https://github.com/ashishpatel26/365-Days-Computer-Vision-Learning-Linkedin-Post
🔸 For this reason, dilated convolutions are used to increase the receptive field of the higher layers, compensating for the reduction in receptive field induced by removing subsampling. 🔸 Compare with original Resnet Final group of convolutional layers g4 and g5 contains 3 x 3 Standard Convolution(d=1) and The feature maps are getting smaller due to the max pooling. The output feature map has the size of 7×7 only. DRN taking (d=2) for same group. Finally, the output of G5 in DRN is 28×28 which is much larger than that of original ResNet.
If someone has done normal Python basics what u suggest go for more project in Python or learn numpy pandas then go into project