If you use this code for a paper please cite:

@article{gao2021container,
  title={Container: Context Aggregation Network},
  author={Gao, Peng and Lu, Jiasen and Li, Hongsheng and Mottaghi, Roozbeh and Kembhavi, Aniruddha},
  journal={arXiv preprint arXiv:2106.01401},
  year={2021}
}

We provide baseline Container-light models pretrained on ImageNet 2012.

First, clone the repository locally:

git clone https://github.com/allenai/container.git

Create a new conda environment:

conda create -n container python=3.7
conda activate container
cd container

Install PyTorch 1.7.0+ and torchvision 0.8.1+ and pytorch-image-models 0.3.2:

conda install -c pytorch pytorch torchvision
pip install timm==0.3.2

Download and extract ImageNet train and val images from http://image-net.org/. The directory structure is the standard layout for the torchvision datasets.ImageFolder, and the training and validation data is expected to be in the train/ folder and val folder respectively:

/path/to/imagenet/
  train/
    class1/
      img1.jpeg
    class2/
      img2.jpeg
  val/
    class1/
      img3.jpeg
    class/2
      img4.jpeg

To evaluate a pre-trained Container-Light on ImageNet val with a single GPU run:

For Container-Light, run:

python main.py --eval --resume checkpoint.pth --model container_v1_light --data-path /path/to/imagenet

giving

* Acc@1 82.26

To train Container-Light on ImageNet on a single node with 8 gpus for 300 epochs run:

Container-Light

python -m torch.distributed.launch --nproc_per_node=8 --use_env main.py --model container_v1_light --batch-size 128 --data-path /path/to/imagenet --output_dir /path/to/save

Code will be released seperately.

Container V2 with much better performance will be released soon. Stay tuned.

Imagenet pretrained model for Container V2 : Container V2

This repository is released under the Apache 2.0 license as found in the LICENSE file.

Container codebase is highly motivated by DeiT