[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-71113":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":18,"rankGlobal":9,"rankLanguage":9,"license":19,"archived":20,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":15,"starSnapshotCount":15,"syncStatus":27,"lastSyncTime":28,"discoverSource":29},71113,"ConvNeXt","facebookresearch\u002FConvNeXt","facebookresearch","Code release for ConvNeXt model",null,"Python",6383,742,6364,51,0,8,21,70.71,"MIT License",true,false,"main",[],"2026-06-12 04:00:59","# [A ConvNet for the 2020s](https:\u002F\u002Farxiv.org\u002Fabs\u002F2201.03545)\n\nOfficial PyTorch implementation of **ConvNeXt**, from the following paper:\n\n[A ConvNet for the 2020s](https:\u002F\u002Farxiv.org\u002Fabs\u002F2201.03545). CVPR 2022.\\\n[Zhuang Liu](https:\u002F\u002Fliuzhuang13.github.io), [Hanzi Mao](https:\u002F\u002Fhanzimao.me\u002F), [Chao-Yuan Wu](https:\u002F\u002Fchaoyuan.org\u002F), [Christoph Feichtenhofer](https:\u002F\u002Ffeichtenhofer.github.io\u002F), [Trevor Darrell](https:\u002F\u002Fpeople.eecs.berkeley.edu\u002F~trevor\u002F) and [Saining Xie](https:\u002F\u002Fsainingxie.com)\\\nFacebook AI Research, UC Berkeley\\\n[[`arXiv`](https:\u002F\u002Farxiv.org\u002Fabs\u002F2201.03545)][[`video`](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=QzCjXqFnWPE)]\n\n--- \n\n\u003Cp align=\"center\">\n\u003Cimg src=\"https:\u002F\u002Fuser-images.githubusercontent.com\u002F8370623\u002F180626875-fe958128-6102-4f01-9ca4-e3a30c3148f9.png\" width=100% height=100% \nclass=\"center\">\n\u003C\u002Fp>\n\nWe propose **ConvNeXt**, a pure ConvNet model constructed entirely from standard ConvNet modules. ConvNeXt is accurate, efficient, scalable and very simple in design.\n\n## Catalog\n- [x] ImageNet-1K Training Code  \n- [x] ImageNet-22K Pre-training Code  \n- [x] ImageNet-1K Fine-tuning Code  \n- [x] Downstream Transfer (Detection, Segmentation) Code\n- [x] Image Classification [\\[Colab\\]](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1CBYTIZ4tBMsVL5cqu9N_-Q3TBprqsfEO?usp=sharing) and Web Demo [![Hugging Face Spaces](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20Hugging%20Face-Spaces-blue)](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fakhaliq\u002Fconvnext)\n- [x] Fine-tune on CIFAR with Weights & Biases logging [\\[Colab\\]](https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1ijAxGthE9RENJJQRO17v9A7PTd1Tei9F?usp=sharing)\n\n\n\n\u003C!-- ✅ ⬜️  -->\n\n## Results and Pre-trained Models\n### ImageNet-1K trained models\n\n| name | resolution |acc@1 | #params | FLOPs | model |\n|:---:|:---:|:---:|:---:| :---:|:---:|\n| ConvNeXt-T | 224x224 | 82.1 | 28M | 4.5G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_tiny_1k_224_ema.pth) |\n| ConvNeXt-S | 224x224 | 83.1 | 50M | 8.7G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_small_1k_224_ema.pth) |\n| ConvNeXt-B | 224x224 | 83.8 | 89M | 15.4G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_1k_224_ema.pth) |\n| ConvNeXt-B | 384x384 | 85.1 | 89M | 45.0G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_1k_384.pth) |\n| ConvNeXt-L | 224x224 | 84.3 | 198M | 34.4G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_large_1k_224_ema.pth) |\n| ConvNeXt-L | 384x384 | 85.5 | 198M | 101.0G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_large_1k_384.pth) |\n\n### ImageNet-22K trained models\n\n| name | resolution |acc@1 | #params | FLOPs | 22k model | 1k model |\n|:---:|:---:|:---:|:---:| :---:| :---:|:---:|\n| ConvNeXt-T | 224x224 | 82.9 | 29M | 4.5G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_tiny_22k_224.pth)   | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_tiny_22k_1k_224.pth)\n| ConvNeXt-T | 384x384 | 84.1 | 29M | 13.1G |     -          | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_tiny_22k_1k_384.pth)\n| ConvNeXt-S | 224x224 | 84.6 | 50M | 8.7G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_small_22k_224.pth)   | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_small_22k_1k_224.pth)\n| ConvNeXt-S | 384x384 | 85.8 | 50M | 25.5G |     -          | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_small_22k_1k_384.pth)\n| ConvNeXt-B | 224x224 | 85.8 | 89M | 15.4G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_22k_224.pth)   | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_22k_1k_224.pth)\n| ConvNeXt-B | 384x384 | 86.8 | 89M | 47.0G |     -          | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_22k_1k_384.pth)\n| ConvNeXt-L | 224x224 | 86.6 | 198M | 34.4G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_large_22k_224.pth)  | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_large_22k_1k_224.pth)\n| ConvNeXt-L | 384x384 | 87.5 | 198M | 101.0G |    -         | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_large_22k_1k_384.pth)\n| ConvNeXt-XL | 224x224 | 87.0 | 350M | 60.9G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_xlarge_22k_224.pth) | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_xlarge_22k_1k_224_ema.pth)\n| ConvNeXt-XL | 384x384 | 87.8 | 350M | 179.0G |  -          | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_xlarge_22k_1k_384_ema.pth)\n\n\n### ImageNet-1K trained models (isotropic)\n| name | resolution |acc@1 | #params | FLOPs | model |\n|:---:|:---:|:---:|:---:| :---:|:---:|\n| ConvNeXt-S | 224x224 | 78.7 | 22M | 4.3G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_iso_small_1k_224_ema.pth) |\n| ConvNeXt-B | 224x224 | 82.0 | 87M | 16.9G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_iso_base_1k_224_ema.pth) |\n| ConvNeXt-L | 224x224 | 82.6 | 306M | 59.7G | [model](https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_iso_large_1k_224_ema.pth) |\n\n\n## Installation\nPlease check [INSTALL.md](INSTALL.md) for installation instructions. \n\n## Evaluation\nWe give an example evaluation command for a ImageNet-22K pre-trained, then ImageNet-1K fine-tuned ConvNeXt-B:\n\nSingle-GPU\n```\npython main.py --model convnext_base --eval true \\\n--resume https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_22k_1k_224.pth \\\n--input_size 224 --drop_path 0.2 \\\n--data_path \u002Fpath\u002Fto\u002Fimagenet-1k\n```\nMulti-GPU\n```\npython -m torch.distributed.launch --nproc_per_node=8 main.py \\\n--model convnext_base --eval true \\\n--resume https:\u002F\u002Fdl.fbaipublicfiles.com\u002Fconvnext\u002Fconvnext_base_22k_1k_224.pth \\\n--input_size 224 --drop_path 0.2 \\\n--data_path \u002Fpath\u002Fto\u002Fimagenet-1k\n```\n\nThis should give \n```\n* Acc@1 85.820 Acc@5 97.868 loss 0.563\n```\n\n- For evaluating other model variants, change `--model`, `--resume`, `--input_size` accordingly. You can get the url to pre-trained models from the tables above. \n- Setting model-specific `--drop_path` is not strictly required in evaluation, as the `DropPath` module in timm behaves the same during evaluation; but it is required in training. See [TRAINING.md](TRAINING.md) or our paper for the values used for different models.\n\n## Training\nSee [TRAINING.md](TRAINING.md) for training and fine-tuning instructions.\n\n## Acknowledgement\nThis repository is built using the [timm](https:\u002F\u002Fgithub.com\u002Frwightman\u002Fpytorch-image-models) library, [DeiT](https:\u002F\u002Fgithub.com\u002Ffacebookresearch\u002Fdeit) and [BEiT](https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Funilm\u002Ftree\u002Fmaster\u002Fbeit) repositories.\n\n## License\nThis project is released under the MIT license. Please see the [LICENSE](LICENSE) file for more information.\n\n## Citation\nIf you find this repository helpful, please consider citing:\n```\n@Article{liu2022convnet,\n  author  = {Zhuang Liu and Hanzi Mao and Chao-Yuan Wu and Christoph Feichtenhofer and Trevor Darrell and Saining Xie},\n  title   = {A ConvNet for the 2020s},\n  journal = {Proceedings of the IEEE\u002FCVF Conference on Computer Vision and Pattern Recognition (CVPR)},\n  year    = {2022},\n}\n```\n","ConvNeXt 是一个纯卷积神经网络模型，完全由标准的卷积网络模块构建而成。该项目提供了 ConvNeXt 模型的官方 PyTorch 实现，具备高精度、高效能和可扩展性，并且设计简洁。核心功能包括在 ImageNet-1K 和 ImageNet-22K 上的预训练及微调代码，以及在下游任务如目标检测和图像分割中的应用示例。此外，还提供了基于 CIFAR 数据集的微调 Colab 笔记本和 Hugging Face 的在线演示。ConvNeXt 适用于需要高性能图像分类和计算机视觉任务的各种场景，尤其是在对模型准确性和效率有较高要求的情况下。",2,"2026-06-11 03:35:57","high_star"]