See In The Dark

This project focuses on Low Light Photo Enhancement Using Simplified Fully Convolutional Neural Networks. This was attempted to produce a memory efficient model that can effectively run on edge devices and deliver an optimum level of image enhancement and noise reduction.

Table of contents

• General Info
• Packages Used
• Setup
• Usage
• Screenshots
• References

General Info

• This project is aimed at creating a solution to tackling the problem of Low Light Photography Using Simplified Fully Convolutional Neural Networks
• The Neural Network model architecture is written in PyTorch and the application is deployed using Flask
• The model was trained on the Sony_gt Dataset which is a subset of the Sony dataset containing images of varying exposures
• This project was done as a part of The MLH Local Hack Day 2019

Packages Used

• PyTorch
• Flask
• requests
• numpy

Setup

Requirements

• Python 3.3+

Installation

• To install required packages:
  

pip install -r requirements.txt

Usage

Linux

export FLASK_APP=app.py
flask run

Windows

set FLASK_APP=app.py
flask run

Screenshots

• Image Upload Page:

• Result Page:

References

• https://github.com/ninetf135246/pytorch-Learning-to-See-in-the-Dark
• https://github.com/cchen156/Learning-to-See-in-the-Dark