{"id":18682539,"url":"https://github.com/postech-cvlab/highqualityframeinterpolation","last_synced_at":"2025-04-12T04:22:11.235Z","repository":{"id":109960673,"uuid":"331636999","full_name":"POSTECH-CVLab/HighQualityFrameInterpolation","owner":"POSTECH-CVLab","description":"An official source code of Choi et al., High-quality Frame Interpolation via Tridirectional Inference, WACV 2021 paper.","archived":false,"fork":false,"pushed_at":"2021-03-26T16:15:21.000Z","size":70106,"stargazers_count":16,"open_issues_count":0,"forks_count":4,"subscribers_count":5,"default_branch":"master","last_synced_at":"2025-03-25T23:51:21.300Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/POSTECH-CVLab.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2021-01-21T13:30:25.000Z","updated_at":"2024-11-21T04:04:27.000Z","dependencies_parsed_at":"2023-03-23T03:02:02.522Z","dependency_job_id":null,"html_url":"https://github.com/POSTECH-CVLab/HighQualityFrameInterpolation","commit_stats":null,"previous_names":[],"tags_count":1,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/POSTECH-CVLab%2FHighQualityFrameInterpolation","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/POSTECH-CVLab%2FHighQualityFrameInterpolation/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/POSTECH-CVLab%2FHighQualityFrameInterpolation/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/POSTECH-CVLab%2FHighQualityFrameInterpolation/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/POSTECH-CVLab","download_url":"https://codeload.github.com/POSTECH-CVLab/HighQualityFrameInterpolation/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248514504,"owners_count":21116973,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","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":[],"created_at":"2024-11-07T10:12:17.029Z","updated_at":"2025-04-12T04:22:11.229Z","avatar_url":"https://github.com/POSTECH-CVLab.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"## High-quality Frame Interpolation via Tridirectional Inference\n\nThis is an official source code of Choi et al., High-quality Frame Interpolation via Tridirectional Inference, WACV 2021 paper.\n\n\u003cp align=\"center\"\u003e\n  \u003cimg width=\"100%\" src=\"https://github.com/POSTECH-CVLab/HighQualityFrameInterpolation/blob/master/figures/overview.jpg?raw=true\" /\u003e\n\u003c/p\u003e\n\n\n## Features\n- Simple and effective video frame interpolation approach using **three consequent video frames**.\n- Data-driven approach to interpolate moving objects.\n- Generalizes well to high-resolution contents.\n\n## Requirements\nWe tested our approach using the below environment. We recommend using docker and anaconda to avoid any hassles. (Tested with 4x Nvidia Titan X (12GB))\n\n- Cuda 10.0\n- cudatoolkit=10.0\n- pytorch==1.0.1\n- torchvision==0.2.2\n- visdom==0.1.8.9\n- [PWCNet](https://github.com/NVlabs/PWC-Net) (included).\n- For more information, please check environment.yml file.\n\n\n## Interpolating your own video with the pre-trained network\n1. Check the file of the pretrained network. ```trained_models/FullNet_v5.pth```\n2. Put your video file. For instance:\n```\ndata/custom/13_1.mp4\n```\n3. Make a image sequence from a video using ```python vid2frames.py```\n4. Run ```run_seq.py```.  For instance:\n```\nCUDA_VISIBLE_DEVICES=0 python run_seq.py --cuda --n_iter 1 --in_file data/custom/13_1/ --out_file data/custom/13_1_out/\n```\n5. Check the output folder to see interpolated frames. ```xx_1.png``` indicates interpolated frames, and ```xx_0.png``` indicates original video frames.\n\n\n## Training the Network\n1. Compile PWCNet in correlation package using the following script: \n```\ncd models/PWCNet/correlation_package_pytorch1_0/ \u0026\u0026 ./build.sh\n``` \n- If compilation is not working, consider modifying ```models/PWCNet/correlation_package_pytorch1_0/setup.py``` by changing the line ```'-gencode', 'arch=compute_75,code=sm_75'``` with the proper version by referencing [this](https://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/).\n\n2. Download [Vimeo-90k Septuplet Dataset](http://data.csail.mit.edu/tofu/dataset/vimeo_septuplet.zip) in ```data``` folder (for providing plentiful video contents).\n3. Download [PASCAL VOC 2012 Dataset](http://host.robots.ox.ac.uk/pascal/VOC/voc2012/VOCtrainval_11-May-2012.tar) in ```data``` folder (for augmenting flying objects).\n4. Uncompress the downloaded files, and organize the data folder like below:\n\n```\n    data\n        ├── vimeo_setuplet\n        │   └── sequences\n        ├── VOCdevkit\n            └── VOC2012\n                ├── :\n                ├── SegmentationObject\n                ├── ObjSegments\n                └── :\n    :\n    train.py\n    :\n```\n5. To add flying objects onto downloaded videos, run ```python seg_obj.py```. It crops image patches of objects from PASCAL VOC and save them into ```data/VOCdevkit/VOC2012/ObjSegments/```\n6. Run Visdom ```python -m visdom.server -port [port_number]```\n7. Train the network. For instance, to use 4 GPUs:\n```\nCUDA_VISIBLE_DEVICES=0,1,2,3 python train.py --cuda --n_epoch 20 --decay_epoch 10 --batchSize 4 --n_cpu 8 --modelname './trained_models/FullNet_v5.pth'\n```\n*consider decrease ```--batchSize``` if the training crashes due to the small GPU memory.\n\n## Trying with other datasets\n(will be updated)\n1. Download [SMBV dataset](http://www.cvg.unibe.ch/media/data/datasets/video/jin/slow-motion.zip) (Utilized four test videos)\n2. Download [GoPro dataset](https://drive.google.com/file/d/1SlURvdQsokgsoyTosAaELc4zRjQz9T2U/view?usp=sharing) (We utilized eleven videos)\n\n## Citation\nPlease cite our work if you use High-quality Frame Interpolation.\n```bib\n@article{Choi2021HQFrameInterpolation,\n  title   = {{High-quality Frame Interpolation via Tridirectional Inference}},\n  author  = {Jinsoo Choi and Jaesik Park and In So Kweon},\n  journal = {Winter Conference on Applications of Computer Vision (WACV)},\n  year    = {2021}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpostech-cvlab%2Fhighqualityframeinterpolation","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fpostech-cvlab%2Fhighqualityframeinterpolation","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpostech-cvlab%2Fhighqualityframeinterpolation/lists"}