DeepDigitalFilm

DigitalFilm: Use a neural network to simulate film style.

---

"DigitalFilm" Digital Film

Use a neural network to simulate film style.
Explore the documentation of this project »

View the demo · Report a bug · Propose a new feature

This README.md is for developers and users 简体中文

PowerPoint Presentation

source code

Table of Contents

• DeepDigitalFilm
  • Table of Contents
    • Sample
    • Run Demo
    • training model - Installation steps
    • Overall architecture
    • Dataset
    • File directory description
    • Version Control
    • Author
    • Copyright

Sample

Figure 1 Sample rollei_infrared_400

Figure 2 Sample kodak gold 200

Figure 3 Sample fuji color 200

Run Demo

The length and width of the input photo need to be divisible by 32.

python digitalFilm.py [-v/-h/-g] -i <input> -o <ouput> -m <model>

• -v print version information
• -h help information
• -g graphical image selection
• -i input image directory
• -o output image directory
• -m model directory

training model

training model directly use cyclegan.ipynb. But you need to download the pre-trained model of resnet18 in advance. Prepare digital photos and film photos in two folders. The model are included in the Release.

Installation steps

git clone https://github.com/SongZihui-sudo/digitalFilm.git

It is best to create an environment in conda now and then install various dependencies.

pip install -r requirement.txt

Overall architecture

Converting digital photos to film style can be regarded as an image style conversion task. Therefore, the overall architecture adopts the cycleGAN network. pytorch-CycleGAN-and-pix2pix In addition, it is difficult to obtain large-scale digital photos and film-style photos, so an unsupervised approach is adopted to use unpaired data for training.

Dataset

The dataset consists of dual-source image data, the main part of which is collected from high-quality digital photos taken by Xiaomi 13 Ultra mobile phone, and the rest is selected from professional HDR image dataset. Film samples are collected from the Internet.

File directory description

• DigitalFilm.ipynb is used to train the model
• app is a demo
• digitalFilm.py
• mynet.py
• mynet2.py

Version Control

This project uses Git for version management. You can view the currently available version in the repository.

Author

151122876@qq.com SongZihui-sudo

Zhihu:Dr.who   qq:1751122876

*You can also view all the developers involved in the project in the list of contributors. *

Copyright

This project is licensed under GPLv3. For details, please refer to LICENSE.txt