https://github.com/imvision12/convnext-tf
A Tensorflow Implementation of "A ConvNet for the 2020s" and "ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders" (ConvNeXt and ConvNeXtV2)
https://github.com/imvision12/convnext-tf
cnn convnext convnextv2 tensorflow
Last synced: 9 months ago
JSON representation
A Tensorflow Implementation of "A ConvNet for the 2020s" and "ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders" (ConvNeXt and ConvNeXtV2)
- Host: GitHub
- URL: https://github.com/imvision12/convnext-tf
- Owner: IMvision12
- License: mit
- Created: 2023-02-11T14:44:44.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2023-02-11T22:22:44.000Z (over 3 years ago)
- Last Synced: 2024-11-14T11:37:56.356Z (over 1 year ago)
- Topics: cnn, convnext, convnextv2, tensorflow
- Language: Python
- Homepage:
- Size: 332 KB
- Stars: 4
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
### To-do
- [x] Add convnextv1 and v2 pytorch
- [x] convert pytorch to tensorflow
- [ ] weight conversion
# ConvNeXt and ConvNeXtV2
This repository is about an implementation of the research paper "A ConvNet of the 2020s" and "ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders" using `Tensorflow`.
ConvNeXtV1 : ConvNeXt, a pure ConvNet model 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.
ConvNeXtV2: The paper proposed a fully convolutional masked autoencoder framework (FCMAE) and a new Global Response Normalization (GRN) layer to original ConvNeXtV1 model 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.
### ConvNeXtV1 and ConvNeXtV2 block design:
# References
[1] ConvNeXt paper: https://arxiv.org/abs/2201.03545
[2] ConvNeXtV2 paper: https://arxiv.org/abs/2301.00808
[3] Official ConvNeXt code: https://github.com/facebookresearch/ConvNeXt
[4] Official ConvNeXtV2 code: https://github.com/facebookresearch/ConvNeXt-V2