Introduction

This simple image viewer is specialized in AI Super Resolution Upscaling in real-time. Using current Super Resolution AI models for improved perceived upscaling quality. It has all the basic image viewing functions and many visual and speed focused implementations. All the internal image scaling are performed by GPU accelerated shaders. Current version contains about 500 lines of Python codes plus external shader files. This project shares its Super Resolution backends with my other repos. Ths project will run cross-platform under both Windows and Linux (under testing).

Features

• GPU accelerated AI SRCNN for high quality 2x or 4x cascaded zoom.
• GPU accelerated Lanczos or Bicubic fit to screen image browsing.
• Pygame(SDL) based cross-platform GUI, highly consistant and responsive.
• Fast image viewing folder by folder.
• Image decodings are prefetched and cached for speed.
• Auto bookmark and resume support.
• PIXIV specific file grouping and batch moving/deleting functions.

Future plans

• Implement other HLSL/GLSL shaders cross-platform using the compushady library.
• Implement other experimental Super Resolution AI models with simple pytorch calls.
• [Next revision][Shader] Improve x4 shaders speed by elimiting unnecessary intermediate buffer I/O.

Known issues

• [Linux] System font spacing/placement and mouse cusor change not working as intended. => use user font/cursors.
• [Linux] Folder/File dialog crashes, after resolution change?

Acknowledgement and Special Thanks

This project contains codes based on the following projects/libraries:

• https://github.com/Blinue/Magpie
• https://github.com/funnyplanter/CuNNy
• https://github.com/rdeioris/compushady
• https://www.pygame.org/