From f73a389fd38ebca0fdf5b180eb6eb3f3fec77e57 Mon Sep 17 00:00:00 2001 From: Rishit Dagli Date: Sun, 20 Dec 2020 23:40:19 +0530 Subject: [PATCH] Add Architecture details --- README.md | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) diff --git a/README.md b/README.md index 4e669c2..bce2ecf 100644 --- a/README.md +++ b/README.md @@ -8,3 +8,21 @@ This repository shows the TensorFlow Lite and TensorRT model conversion and infe

Model training code and pre-trained weights are provided by Soumik through [this repository](https://github.com/soumik12345/MIRNet/). + +## A bit about the architecture + +The **MIRNet** presents a novel architecture with the collective goals of maintaining high-resolution representations through the entire network, and +receiving strong contextual information from the low-resolution representations. + +The core of this approach is a multi-scale residual block containing the following key elements: +- parallel multi-resolution convolution streams for extracting multi-scale features +- information exchange across the multi-resolution streams +- spatial and channel attention mechanisms for capturing contextual information +- attention based multi-scale feature aggregation. + +![](https://i.imgur.com/VOzfG9a.png) +

+Framework of MIRNet +

+ +The figure above shows the framework of the proposed network MIRNet that learns enriched feature representations for image restoration and enhancement.