README_en.md

Smartinput-Upscaler

(中文 | English)

Initially, it was because the school's ancestral ppt was too blurry, and I wanted to try to make it clearer. The project can run on both Windows and Linux.

The project is based on Real-ESRGAN for convenient and fast super-resolution processing of various images/videos.

GUI used PyQt-Fluent-Widgets.

TODO List

• (Almost) lossless compression after image enlargement.
• Add video support.
• Implement batch processing for images.

Usage

Currently, image and PDF inputs are supported. You can download the corresponding system version from Releases on the right side, which should be ready to use out of the box.

Note that if the source PDF is already clear enough, it may have a counterproductive effect! (i.e., even less clear)

In most cases, the conversion effect of PDF is as follows:

Multi-platform support

Model and File Selection

Since the backend uses Real-ESRGAN-ncnn-vulkan, it is able to use 3 models and supports Intel/AMD/Nvidia graphics card acceleration. The default model is the one suitable for general-purpose image processing.

Progress bar with remaining time estimate

Building from Source Code

Install python environment

conda create -n pdf_up python=3.11
conda activate pdf_up
pip install PyMuPDF
pip install pyqt6
pip install PyQt6-Fluent-Widgets -i https://pypi.org/simple/
python GUI.py

The model used comes from the Real-ESRGAN project. Please download the executable file from below and extract it in the root directory.

Real-ESRGAN_README