convnext

ConvNeXt

A ConvNet for the 2020s

Introduction

ConvNeXt is initially described in A ConvNet for the 2020s, which is a pure convolutional model (ConvNet), inspired by the design of Vision Transformers. The ConvNeXt has the pyramid structure and achieve competitive performance on various vision tasks, with simplicity and efficiency.

Abstract

Show the paper's abstract

The "Roaring 20s" of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually "modernize" a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets.

How to use it?

Predict image

from mmpretrain import inference_model

predict = inference_model('convnext-tiny_32xb128_in1k', 'demo/bird.JPEG')
print(predict['pred_class'])
print(predict['pred_score'])

Use the model

import torch
from mmpretrain import get_model

model = get_model('convnext-tiny_32xb128_in1k', pretrained=True)
inputs = torch.rand(1, 3, 224, 224)
out = model(inputs)
print(type(out))
# To extract features.
feats = model.extract_feat(inputs)
print(type(feats))

Train/Test Command

Prepare your dataset according to the docs.

Train:

python tools/train.py configs/convnext/convnext-tiny_32xb128_in1k.py

Test:

python tools/test.py configs/convnext/convnext-tiny_32xb128_in1k.py https://download.openmmlab.com/mmclassification/v0/convnext/convnext-tiny_32xb128_in1k_20221207-998cf3e9.pth

Models and results

Pretrained models

Model	Params (M)	Flops (G)	Config	Download
convnext-base_3rdparty_in21k*	88.59	15.36	config	model
convnext-large_3rdparty_in21k*	197.77	34.37	config	model
convnext-xlarge_3rdparty_in21k*	350.20	60.93	config	model

Models with * are converted from the official repo. The config files of these models are only for inference. We haven't reproduce the training results.

Image Classification on ImageNet-1k

Model	Pretrain	Params (M)	Flops (G)	Top-1 (%)	Top-5 (%)	Config	Download
convnext-tiny_32xb128_in1k	From scratch	28.59	4.46	82.14	96.06	config	model | log
convnext-tiny_32xb128-noema_in1k	From scratch	28.59	4.46	81.95	95.89	config	model | log
convnext-tiny_in21k-pre_3rdparty_in1k*	ImageNet-21k	28.59	4.46	82.90	96.62	config	model
convnext-tiny_in21k-pre_3rdparty_in1k-384px*	ImageNet-21k	28.59	13.14	84.11	97.14	config	model
convnext-small_32xb128_in1k	From scratch	50.22	8.69	83.16	96.56	config	model | log
convnext-small_32xb128-noema_in1k	From scratch	50.22	8.69	83.21	96.48	config	model | log
convnext-small_in21k-pre_3rdparty_in1k*	ImageNet-21k	50.22	8.69	84.59	97.41	config	model
convnext-small_in21k-pre_3rdparty_in1k-384px*	ImageNet-21k	50.22	25.58	85.75	97.88	config	model
convnext-base_32xb128_in1k	From scratch	88.59	15.36	83.66	96.74	config	model | log
convnext-base_32xb128-noema_in1k	From scratch	88.59	15.36	83.64	96.61	config	model | log
convnext-base_3rdparty_in1k*	From scratch	88.59	15.36	83.85	96.74	config	model
convnext-base_3rdparty-noema_in1k*	From scratch	88.59	15.36	83.71	96.60	config	model
convnext-base_3rdparty_in1k-384px*	From scratch	88.59	45.21	85.10	97.34	config	model
convnext-base_in21k-pre_3rdparty_in1k*	ImageNet-21k	88.59	15.36	85.81	97.86	config	model
convnext-base_in21k-pre-3rdparty_in1k-384px*	From scratch	88.59	45.21	86.82	98.25	config	model
convnext-large_3rdparty_in1k*	From scratch	197.77	34.37	84.30	96.89	config	model
convnext-large_3rdparty_in1k-384px*	From scratch	197.77	101.10	85.50	97.59	config	model
convnext-large_in21k-pre_3rdparty_in1k*	ImageNet-21k	197.77	34.37	86.61	98.04	config	model
convnext-large_in21k-pre-3rdparty_in1k-384px*	From scratch	197.77	101.10	87.46	98.37	config	model
convnext-xlarge_in21k-pre_3rdparty_in1k*	ImageNet-21k	350.20	60.93	86.97	98.20	config	model
convnext-xlarge_in21k-pre-3rdparty_in1k-384px*	From scratch	350.20	179.20	87.76	98.55	config	model

Models with * are converted from the official repo. The config files of these models are only for inference. We haven't reproduce the training results.

Citation

@Article{liu2022convnet,
  author  = {Zhuang Liu and Hanzi Mao and Chao-Yuan Wu and Christoph Feichtenhofer and Trevor Darrell and Saining Xie},
  title   = {A ConvNet for the 2020s},
  journal = {arXiv preprint arXiv:2201.03545},
  year    = {2022},
}