🎯Generative Refocusing is out🎯 👉Generative Refocusing is a two-step process that uses...

2025-12-23 · Source: AI with Papers - Artificial Intelligence & Deep Learning (@AI_DeepLearning) - Telegram · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, quick

Summary

Generative Refocusing is a new two-step image processing technique designed to produce all-in-focus images and controllable bokeh effects. The process utilizes DeblurNet to reconstruct sharp images from various input types, followed by BokehNet, which operates in a semi-supervised mode to generate customizable bokeh. This project's code is released under the Apache 2.0 license, making it openly accessible. Resources including a review, the research paper (arxiv.org/pdf/2512.16923), a dedicated project page (generative-refocusing.github.io/), a GitHub repository (github.com/rayray9999/Genfocus), and a Hugging Face demo (huggingface.co/spaces/nycu-cplab/Genfocus-Demo) are available for further exploration.

Key takeaway

For Computer Vision Engineers or AI Scientists developing image enhancement tools, Generative Refocusing offers a robust, open-source solution for deblurring and artistic bokeh effects. You should explore its DeblurNet and semi-supervised BokehNet components to integrate advanced refocusing capabilities into your applications, potentially improving visual quality and creative control in image processing pipelines. The Apache 2.0 license allows for flexible adoption.

Key insights

Generative Refocusing uses a two-step neural network process for deblurring and controllable bokeh generation.

Principles

• Two-step process for image refocusing
• Semi-supervised bokeh generation

Method

The method involves DeblurNet for all-in-focus image recovery from diverse inputs, then BokehNet for controllable bokeh creation in a semi-supervised manner.

In practice

• Recover sharp images from blurred inputs
• Add customizable bokeh to images

Topics

• Generative Refocusing
• Image Processing
• DeblurNet
• BokehNet
• Semi-supervised Learning

Best for: Computer Vision Engineer, AI Scientist, Research Scientist, AI Researcher, AI Engineer, Deep Learning Engineer

Related on AIssential

• How this AI Learned to Refocus Any Photo — AI refocusing allows dynamic post-capture adjustment of photographic focus and depth of field from a single image.
• Towards Photorealistic and Efficient Bokeh Rendering via Diffusion Framework — MagicBokeh unifies bokeh rendering and super-resolution using a diffusion framework for photorealistic effects on low-resolution images.
• FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining — FreeStyle uses community LoRAs and a two-stage curriculum for robust style-content dual-reference image generation.
• OmniDirector: General Multi-Shot Camera Cloning without Cross-Paired Data — The OmniDirector framework enables general multi-shot camera cloning by encoding camera motion as grid videos and training on million-scale data.
• PhyEditBench: A Real-World Multi-Stage Benchmark for Physics-Aware Image Editing — Benchmarking physics-aware image editing reveals current model limitations and suggests video generation as a reasoning mechanism.
• Rethinking Token Reduction for Diffusion Models via Output-Similarity-Awareness — DiTo improves Diffusion Transformer efficiency by focusing token reduction on output similarity, not just input similarity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI with Papers - Artificial Intelligence & Deep Learning (@AI_DeepLearning) - Telegram.