AI Rank
← Open Source
facebookresearch

ConvNeXt-V2

Code release for ConvNeXt V2 model.

ModelsFoundation modelsVision modelsPython
Open on GitHub
Momentum
+0stars in 24 hours0.0%
2.06k
Stars
177
Forks
+-2
This week
3
Contributors
Created 2022-12-29 · Updated 2026-10-12 · #9030 today
Top developers
sdebnathuscs9xiesorrowyn

facebookresearch/ConvNeXt-V2 is ranked 9,030 with 2,065 stars.

Same-category comparison

facebookresearch/ConvNeXt-V2 is ranked 9,030 among these 8 peers; cfzd/Ultra-Fast-Lane-Detection is ranked 9,037.

Same-category comparison
NameRankMetrics
SUDO-AI-3D/zero123plus8,983
Language:
Python
Stars:
2,102
Forks:
149
Description:
Code repository for Zero123++: a Single Image to Consistent Multi-view Diffusion Base Model.
facebookresearch/chameleon8,984
Language:
Python
Stars:
2,101
Forks:
117
Description:
Repository for Meta Chameleon, a mixed-modal early-fusion foundation model from FAIR.
ai-forever/ru-gpts9,003
Language:
Python
Stars:
2,089
Forks:
433
Description:
Russian GPT3 models.
bytedance/Protenix9,012
Language:
Python
Stars:
2,080
Forks:
328
Description:
Toward High-Accuracy Open-Source Biomolecular Structure Prediction.
cfzd/Ultra-Fast-Lane-Detection9,037
Language:
Python
Stars:
2,063
Forks:
524
Description:
Ultra Fast Structure-aware Deep Lane Detection (ECCV 2020).
serengil/retinaface9,041
Language:
Python
Stars:
2,060
Forks:
201
Description:
RetinaFace: Deep Face Detection Library for Python.
TachibanaYoshino/AnimeGANv39,056
Language:
Python
Stars:
2,044
Forks:
267
Description:
Use AnimeGANv3 to make your own animation works, including turning photos or videos into anime.
alibaba/AliceMind9,066
Language:
Python
Stars:
2,038
Forks:
301
Description:
ALIbaba's Collection of Encoder-decoders from MinD (Machine IntelligeNce of Damo) Lab.
Key parameters and specifications
Key parameters and specifications
ParameterValue
LanguagePython
CategoryModels
SubcategoryFoundation models, Vision models
Stars2,065
Forks177
Contributors3
Created date2022-12-29
Updated date2026-10-12
Stars in 1 day0
Stars in 7 days-2
Rank9,030
Previous rank9,069
Rank change39
7-day stars rank18,246
Previous 7-day stars rank18,199
7-day stars rank change-47
Total stars rank6,176
Previous total stars rank6,168
Total stars rank change-8
CountryUSA
README summary

ConvNeXt V2 Official PyTorch Implementation This repo contains the PyTorch version of *8* model definitions (*Atto, Femto, Pico, Nano, Tiny, Base, Large, Huge*), pre-training/fine-tuning code and pre-trained weights (converted from JAX weights trained on TPU) for our ConvNeXt V2 paper. **ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders** Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon and Saining Xie\ KAIST, Meta AI and New York University We propose a

fully convolutional masked autoencoder framework (FCMAE) and a new Global Response Normalization (GRN) layer that can be added to the ConvNeXt architecture to enhance inter-channel feature competition. This co-design of self-supervised learning techniques and architectural improvement results in a new model family called ConvNeXt V2, which significantly improves the performance of pure ConvNets on various recognition benchmarks. We also provide pre-trained ConvNeXt V2 models of various sizes. 7M | model | | ConvNeXt V2-F | 224x224 | 5. 2M | model

| | ConvNeXt V2-P | 224x224 | 9. 1M | model | | ConvNeXt V2-N | 224x224 | 15. 6M| model | | ConvNeXt V2-T | 224x224 | 28. 55G | model | | ConvNeXt V2-F | 224x224 | 78. 78G | model | | ConvNeXt V2-P | 224x224 | 80. 37G | model | | ConvNeXt V2-N | 224x224 | 81. 45G | model | | ConvNeXt V2-T | 224x224 | 83. 47G | model | | ConvNeXt V2-B | 224x224 | 84. 4G | model | | ConvNeXt V2-L | 224x224 | 85. 4G | model | | ConvNeXt V2-H | 224x224 | 86. 45G | model| | ConvNeXt V2-N | 384x384 | 83. 21G | model| | ConvNeXt V2-T

| 224x224 | 83. 47G | model| | ConvNeXt V2-T | 384x384 | 85. 1G | model| | ConvNeXt V2-B | 224x224 | 86. 4G | model| | ConvNeXt V2-B | 384x384 | 87. 2G | model| | ConvNeXt V2-L | 224x224 | 87. 4G | model| | ConvNeXt V2-L | 384x384 | 88. 1G | model| | ConvNeXt V2-H | 384x384 | 88. 9G | model| | ConvNeXt V2-H | 512x512 | 88. 8G | model| Installation Please check INSTALL. Evaluation We provide example evaluation commands for ConvNeXt V2-Base: Single-GPU Multi-GPU For evaluating other model variants, change , , accordingly. URLs for

the pre-trained models can be found from the result tables. Setting model-specific is not strictly required in evaluation, as the module in timm behaves the same during evaluation; but it is required in training. md or our paper (appendix) for the values used for different models. md for pre-training and fine-tuning instructions. Acknowledgement This repository borrows from timm, ConvNeXt and MAE. We thank Ross Wightman for the initial design of the small-compute ConvNeXt model variants

and the associated training recipe. We also appreciate the helpful discussions and feedback provided by Kaiming He. License This project is released under the MIT license except ImageNet pre-trained and fine-tuned models which are licensed under a CC-BY-NC. Please see the LICENSE file for more information. Citation If you find this repository helpful, please consider citing:

Frequently asked questions

How many stars and forks are recorded?

facebookresearch/ConvNeXt-V2 has 2,065 stars and 177 forks.

What is the repository rank?

facebookresearch/ConvNeXt-V2 is ranked 9,030.

What language and category are recorded?

facebookresearch/ConvNeXt-V2 has 2,065 stars, uses Python and belongs to Models.

Which repository is closest in rank?

facebookresearch/ConvNeXt-V2 is compared with cfzd/Ultra-Fast-Lane-Detection, ranked 9,037 with 2,063 stars.

README

ConvNeXt V2

Official PyTorch Implementation

This repo contains the PyTorch version of 8 model definitions (Atto, Femto, Pico, Nano, Tiny, Base, Large, Huge), pre-training/fine-tuning code and pre-trained weights (converted from JAX weights trained on TPU) for our ConvNeXt V2 paper.

ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders

Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon and Saining Xie\

KAIST, Meta AI and New York University

We propose a fully convolutional masked autoencoder framework (FCMAE) and a new Global Response Normalization (GRN) layer that can be added to the ConvNeXt architecture to enhance inter-channel feature competition. This co-design of self-supervised learning techniques and architectural improvement results in a new model family called ConvNeXt V2, which significantly improves the performance of pure ConvNets on various recognition benchmarks. We also provide pre-trained ConvNeXt V2 models of various sizes.

Results and Pre-trained Models

ImageNet-1K FCMAE pre-trained weights (self-supervised)

name resolution #params model
ConvNeXt V2-A 224x224 3.7M model
ConvNeXt V2-F 224x224 5.2M model
ConvNeXt V2-P 224x224 9.1M model
ConvNeXt V2-N 224x224 15.6M model
ConvNeXt V2-T 224x224 28.6M model
ConvNeXt V2-B 224x224 89M model
ConvNeXt V2-L 224x224 198M model
ConvNeXt V2-H 224x224 660M model

ImageNet-1K fine-tuned models

name resolution acc@1 #params FLOPs model
ConvNeXt V2-A 224x224 76.7 3.7M 0.55G model
ConvNeXt V2-F 224x224 78.5 5.2M 0.78G model
ConvNeXt V2-P 224x224 80.3 9.1M 1.37G model
ConvNeXt V2-N 224x224 81.9 15.6M 2.45G model
ConvNeXt V2-T 224x224 83.0 28.6M 4.47G model
ConvNeXt V2-B 224x224 84.9 89M 15.4G model
ConvNeXt V2-L 224x224 85.8 198M 34.4G model
ConvNeXt V2-H 224x224 86.3 660M 115G model

ImageNet-22K fine-tuned models

name resolution acc@1 #params FLOPs model
ConvNeXt V2-N 224x224 82.1 15.6M 2.45G model
ConvNeXt V2-N 384x384 83.4 15.6M 7.21G model
ConvNeXt V2-T 224x224 83.9 28.6M 4.47G model
ConvNeXt V2-T 384x384 85.1 28.6M 13.1G model
ConvNeXt V2-B 224x224 86.8 89M 15.4G model
ConvNeXt V2-B 384x384 87.7 89M 45.2G model
ConvNeXt V2-L 224x224 87.3 198M 34.4G model
ConvNeXt V2-L 384x384 88.2 198M 101.1G model
ConvNeXt V2-H 384x384 88.7 660M 337.9G model
ConvNeXt V2-H 512x512 88.9 660M 600.8G model

Installation

Please check INSTALL.md for installation instructions.

Evaluation

We provide example evaluation commands for ConvNeXt V2-Base:

Single-GPU

python main_finetune.py \
--model convnextv2_base \
--eval true \
--resume /path/to/checkpoint \
--input_size 224 \
--data_path /path/to/imagenet-1k \

Multi-GPU

python -m torch.distributed.launch --nproc_per_node=8 main_finetune.py \
--model convnextv2_base \
--eval true \
--resume /path/to/checkpoint \
--input_size 224 \
--data_path /path/to/imagenet-1k \
  • For evaluating other model variants, change --model, --resume, --input_size accordingly. URLs for the pre-trained models can be found from the result tables.
  • 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 or our paper (appendix) for the values used for different models.

Training

See TRAINING.md for pre-training and fine-tuning instructions.

Acknowledgement

This repository borrows from timm, ConvNeXt and MAE.

We thank Ross Wightman for the initial design of the small-compute ConvNeXt model variants and the associated training recipe. We also appreciate the helpful discussions and feedback provided by Kaiming He.

License

This project is released under the MIT license except ImageNet pre-trained and fine-tuned models which are licensed under a CC-BY-NC. Please see the LICENSE file for more information.

Citation

If you find this repository helpful, please consider citing:

@article{Woo2023ConvNeXtV2,
  title={ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders},
  author={Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xinlei Chen, Zhuang Liu, In So Kweon and Saining Xie},
  year={2023},
  journal={arXiv preprint arXiv:2301.00808},
}
Nearby in the ranking
Similar items