{"id":43770866,"url":"https://github.com/timmh/neural-supersampling","last_synced_at":"2026-02-05T16:35:44.408Z","repository":{"id":53701353,"uuid":"518003326","full_name":"timmh/neural-supersampling","owner":"timmh","description":"Unofficial re-implementation of a neural supersampling model for real-time rendering","archived":false,"fork":false,"pushed_at":"2024-01-30T14:09:00.000Z","size":30,"stargazers_count":29,"open_issues_count":0,"forks_count":7,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-12-02T14:57:20.500Z","etag":null,"topics":["blender","computer-graphics","deep-learning","super-resolution","supersampling"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/timmh.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2022-07-26T09:50:55.000Z","updated_at":"2025-05-27T08:39:04.000Z","dependencies_parsed_at":"2022-09-12T11:21:46.206Z","dependency_job_id":null,"html_url":"https://github.com/timmh/neural-supersampling","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/timmh/neural-supersampling","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/timmh%2Fneural-supersampling","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/timmh%2Fneural-supersampling/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/timmh%2Fneural-supersampling/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/timmh%2Fneural-supersampling/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/timmh","download_url":"https://codeload.github.com/timmh/neural-supersampling/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/timmh%2Fneural-supersampling/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29125908,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-05T14:05:12.718Z","status":"ssl_error","status_checked_at":"2026-02-05T14:03:53.078Z","response_time":65,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["blender","computer-graphics","deep-learning","super-resolution","supersampling"],"created_at":"2026-02-05T16:35:44.353Z","updated_at":"2026-02-05T16:35:44.402Z","avatar_url":"https://github.com/timmh.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Neural Supersampling\n\nThis is a work-in-progress unofficial re-implementation of the real-time neural supersampling model proposed in `Neural supersampling for real-time rendering` [[`Paper`](https://dl.acm.org/doi/10.1145/3386569.3392376)] using [PyTorch](https://pytorch.org/) and [PyTorch Lightning](https://lightning.ai/). This is in no way endorsed by the original authors.\n\n## Differences\nThis model is implemented as closely to the original paper as possible. However, there are some important differences:\n\n- the original training data is not freely available. Therefore [Blender](https://www.blender.org/) is used to render images with color, depth and motion data from [Blender Open Movies](https://studio.blender.org/films/).\n- the original paper seems to use motion data of the target resolution. Here, due to storage constraints, we use motion data of the source resolution\n- the original paper seems to use raw depth values for feature extraction. I found high depth values to negatively impact numerical stability and therefore decided to use inverse depth, i.e. disparity, instead.\n\n## Rendering\nThe training data may be rendered by Blender and the Cycles rendering engine. To achieve this, download any number of [Blender Open Movie](https://studio.blender.org/films/) assets and configure them in [render_all.py](rendering/render_all.py). Then either run [render_all.py](rendering/render_all.py) directly or use [run_blender_headless.sh](rendering/run_blender_headless.sh) to run Blender via Docker.\n\n## Training\nThe training, evaluation and visualization are all implemented as separate files in the [model](model) directory. Alternatively, take a look at the Jupyter Notebook [NeuralSupersampling.ipynb](NeuralSupersampling.ipynb) \u003ca href=\"https://colab.research.google.com/github/timmh/neural-supersampling/blob/main/NeuralSupersampling.ipynb\"\u003e\u003cimg src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open in Colab\"/\u003e\u003c/a\u003e\n\n## TODO\n- [ ] train to convergence\n- [ ] optimize using [TensorRT](https://github.com/pytorch/TensorRT) and embed in real-time application","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftimmh%2Fneural-supersampling","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Ftimmh%2Fneural-supersampling","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Ftimmh%2Fneural-supersampling/lists"}