Python implementation of the paper with slight changes
Budget: ₹12,500 – ₹37,500 INR
The attached paper uses EMA for feature aggregation. I am looking for a coder who can use SMEMA (split-and-merge EMA) instead.
The other change I am hoping is the use of Res-UTNet50 backbone. The mentioned backbone can be found at : https://github.com/yhygao/UTNet
The joint learning and dense fusion part of the project can be found at: https://github.com/jiangyao-scu/JL-DCF-pytorch
The paper uses squeeze and excitation blocks only at the last block. When the above two are finished, I would like to see how performance varies if the blocks are used alternatively.
A rough idea can be found in JL.pdf.
Rest of the files have self-explanatory file names
The other change I am hoping is the use of Res-UTNet50 backbone. The mentioned backbone can be found at : https://github.com/yhygao/UTNet
The joint learning and dense fusion part of the project can be found at: https://github.com/jiangyao-scu/JL-DCF-pytorch
The paper uses squeeze and excitation blocks only at the last block. When the above two are finished, I would like to see how performance varies if the blocks are used alternatively.
A rough idea can be found in JL.pdf.
Rest of the files have self-explanatory file names