Implementation of paper(python: pytorch) -- 2
Budget: ₹12,500 – ₹37,500 INR
I am looking for a implementation of the enclosed joint learning paper (Looking_at_Boundary) with a backbone of ResUTNet-50 instead of the convolutional backbone. I believe reference for the backbone can be found here: https://github.com/yhygao/UTNet
The joint learning architecture is implemented here: https://github.com/jiangyao-scu/JL-DCF-pytorch
The major part of the work is involved in the feature aggregation module. The paper uses Expectation Maximization Attention algorithm based on Expectation Maximization. My requirement is for Split and Merge Expectation Maximization Attention be used based on the enclosed paper on Split and Merge Expectation Maximization (SMEM Algorithm)
The paper uses a single Squeeze and Excitation block after the convolutional block. I would like a demonstration of how performance changes if the Squeeze and Excitation blocks are used alternatively throughout the ResUTNet-50 backbone please.
The joint learning architecture is implemented here: https://github.com/jiangyao-scu/JL-DCF-pytorch
The major part of the work is involved in the feature aggregation module. The paper uses Expectation Maximization Attention algorithm based on Expectation Maximization. My requirement is for Split and Merge Expectation Maximization Attention be used based on the enclosed paper on Split and Merge Expectation Maximization (SMEM Algorithm)
The paper uses a single Squeeze and Excitation block after the convolutional block. I would like a demonstration of how performance changes if the Squeeze and Excitation blocks are used alternatively throughout the ResUTNet-50 backbone please.