We propose MogaNet, a new family of efficient ConvNets designed through the lens of multi-order game-theoretic interaction, to pursue informative context mining with preferable complexity-performance trade-offs. It shows excellent scalability and attains competitive results among state-of-the-art models with more efficient use of model parameters on ImageNet and multifarious typical vision benchmarks, including COCO object detection, ADE20K semantic segmentation, 2D\&3D human pose estimation, and video prediction.

This repository contains PyTorch implementation for MogaNet (ICLR 2024).

Catalog

We plan to release implementations of MogaNet in a few months. Please watch us for the latest release. Currently, this repo is reimplemented according to our official implementations in OpenMixup, and we are working on cleaning up experimental results and code implementations. Models are released in GitHub / Baidu Cloud / Hugging Face.

• [x] ImageNet-1K Training and Validation Code with timm [code] [models] [Hugging Face 🤗]
• [x] ImageNet-1K Training and Validation Code in OpenMixup / MMPretrain (TODO)
• [x] Downstream Transfer to Object Detection and Instance Segmentation on COCO [code] [models] [demo]
• [x] Downstream Transfer to Semantic Segmentation on ADE20K [code] [models] [demo]
• [x] Downstream Transfer to 2D Human Pose Estimation on COCO [code] (baselines supported) [models] [demo]
  • [ ] Downstream Transfer to 3D Human Pose Estimation (baseline models will be supported)
• [x] Downstream Transfer to Video Prediction on MMNIST Variants [code] (baselines supported)
• [x] Image Classification on Google Colab and Notebook Demo [demo]

Image Classification

1. Installation

Please check INSTALL.md for installation instructions.

2. Training and Validation

See TRAINING.md for ImageNet-1K training and validation instructions, or refer to our OpenMixup implementations. We released pre-trained models on OpenMixup in moganet-in1k-weights. We have also reproduced ImageNet results with this repo and released args.yaml / summary.csv / model.pth.tar in moganet-in1k-weights. The parameters in the trained model can be extracted by code.

Here is a notebook demo of MogaNet which run the steps to perform inference with MogaNet for image classification.

3. ImageNet-1K Trained Models

4. Analysis Tools

(1) The code to count MACs of MogaNet variants.

python get_flops.py --model moganet_tiny

(2) The code to visualize Grad-CAM activation maps (or variants of Grad-CAM) of MogaNet and other popular architectures.

python cam_image.py --use_cuda --image_path /path/to/image.JPEG --model moganet_tiny --method gradcam

(back to top)

5. Downstream Tasks

License

This project is released under the Apache 2.0 license.

Acknowledgement

Our implementation is mainly based on the following codebases. We gratefully thank the authors for their wonderful works.

• pytorch-image-models (timm): PyTorch image models, scripts, pretrained weights.
• PoolFormer: Official PyTorch implementation of MetaFormer.
• ConvNeXt: Official PyTorch implementation of ConvNeXt.
• OpenMixup: Open-source toolbox for visual representation learning.
• MMDetection: OpenMMLab Detection Toolbox and Benchmark.
• MMSegmentation: OpenMMLab Semantic Segmentation Toolbox and Benchmark.
• MMPose: OpenMMLab Pose Estimation Toolbox and Benchmark.
• MMHuman3D: OpenMMLab 3D Human Parametric Model Toolbox and Benchmark.
• OpenSTL: A Comprehensive Benchmark of Spatio-Temporal Predictive Learning.

Citation

If you find this repository helpful, please consider citing:

@inproceedings{iclr2024MogaNet,
  title={MogaNet: Multi-order Gated Aggregation Network},
  author={Siyuan Li and Zedong Wang and Zicheng Liu and Cheng Tan and Haitao Lin and Di Wu and Zhiyuan Chen and Jiangbin Zheng and Stan Z. Li},
  booktitle={International Conference on Learning Representations},
  year={2024}
}

(back to top)