Attention-guided Chained Context Aggregation for Semantic Segmentation
1)CANet is a Encode Decode based semantic segmentation model. Link: https://arxiv.org/pdf/2002.12041.pdf
2)The data set used was Indian Driving Dataset Link: https://idd.insaan.iiit.ac.in/
2)The model use Dilated Resnet 26 as backbone network.
3)Model gains features of various spatial scales through chain-connected ladder-style information flows.
4)Here the Train_Model file contain the train code ,Test_Model contains test and Preprocessing file containe the preprocessing of data.
5)Model final accuracy of Accuracy is 74.06%; Mean IoU as 51.50 and Dice Coefficient as 66.24 on my test dataset.