Home

Awesome

Multiscale Deep Equilibrium Models

:boom::boom::boom::boom:

This repo is deprecated and we will soon stop actively maintaining it, as a more up-to-date (and simpler & more efficient) implementation of MDEQ with the same set of tasks as here is now available in the DEQ repo.

We STRONGLY recommend using with the MDEQ-Vision code in the DEQ repo (which also supports Jacobian-related analysis).

:boom::boom::boom::boom:


This repository contains the code for the multiscale deep equilibrium (MDEQ) model proposed in the paper Multiscale Deep Equilibrium Models by Shaojie Bai, Vladlen Koltun and J. Zico Kolter.

Is implicit deep learning relevant for general, large-scale pattern recognition tasks? We propose the multiscale deep equilibrium (MDEQ) model, which expands upon the DEQ formulation substantially to introduce simultaneous equilibrium modeling of multiple signal resolutions. Specifically, MDEQ solves for and backpropagates through synchronized equilibria of multiple feature representation streams. Such structure rectifies one of the major drawbacks of DEQ, and provide natural hierarchical interfaces for auxiliary losses and compound training procedures (e.g., pretraining and finetuning). Our experiment demonstrate for the first time that "shallow" implicit models can scale to and achieve near-SOTA results on practical computer vision tasks (e.g., megapixel images on Cityscapes segmentation).

We provide in this repo the implementation and the links to the pretrained classification & segmentation MDEQ models.

If you find thie repository useful for your research, please consider citing our work:

@inproceedings{bai2020multiscale,
    author    = {Shaojie Bai and Vladlen Koltun and J. Zico Kolter},
    title     = {Multiscale Deep Equilibrium Models},
    booktitle   = {Advances in Neural Information Processing Systems (NeurIPS)},
    year      = {2020},
}

Overview

The structure of a multiscale deep equilibrium model (MDEQ) is shown below. All components of the model are shown in this figure (in practice, we use n=4).

<p align="center"/> <img src='images/mdeq_structure.png' width="800px"/>

Examples

Some examples of MDEQ segmentation results on the Cityscapes dataset.

<p align="center"/> <img src='images/demo_overlay.gif' width="500px"/> <img src='images/segmentation_example.png' width="800px"/>

Requirements

PyTorch >=1.4.0, torchvision >= 0.4.0

Datasets

All datasets should be downloaded, processed and put in the respective data/[DATASET_NAME] directory. The data/ directory should look like the following:

data/
  cityscapes/
  imagenet/
  ...          (other datasets)
  list/        (see above)

Usage

All experiment settings are provided in the .yaml files under the experiments/ folder.

To train an MDEQ classification model on ImageNet/CIFAR-10, do

python tools/cls_train.py --cfg experiments/[DATASET_NAME]/[CONFIG_FILE_NAME].yaml

To train an MDEQ segmentation model on Cityscapes, do

python -m torch.distributed.launch --nproc_per_node=4 tools/seg_train.py --cfg experiments/[DATASET_NAME]/[CONFIG_FILE_NAME].yaml

where you should provide the pretrained ImageNet model path in the corresponding configuration (.yaml) file. We provide a sample pretrained model extractor in pretrained_models/, but you can also write your own script.

Similarly, to test the model and generate segmentation results on Cityscapes, do

python tools/seg_test.py --cfg experiments/[DATASET_NAME]/[CONFIG_FILE_NAME].yaml

You can (and probably should) initiate the Cityscapes training with an ImageNet-pretrained MDEQ. You need to extract the state dict from the ImageNet checkpointed model, and set the MODEL.PRETRAINED entry in Cityscapes yaml file to this state dict on disk.

The model implementation and MDEQ's algorithmic components (e.g., L-Broyden's method) can be found in lib/.

Pre-trained Models

We provide some reasonably good pre-trained weights here so that one can quickly play with DEQs without training from scratch.

DescriptionTaskDatasetModel
MDEQ-XLImageNet ClassificationImageNetdownload (.pkl)
MDEQ-XLCityscapes(val) SegmentationCityscapesdownload (.pkl)
MDEQ-SmallImageNet ClassificationImageNetdownload (.pkl)
MDEQ-SmallCityscapes(val) SegmentationCityscapesdownload (.pkl)

I. Example of how to evaluate the pretrained ImageNet model:

  1. Download the pretrained ImageNet .pkl file. (I recommend using the gdown command!)
  2. Put the model under pretrained_models/ folder with some file name [FILENAME].
  3. Run the MDEQ classification validation command:
python tools/cls_valid.py --testModel pretrained_models/[FILENAME] --cfg experiments/imagenet/cls_mdeq_[SIZE].yaml

For example, for MDEQ-Small, you should get >75% top-1 accuracy.

II. Example of how to use the pretrained ImageNet model to train on Cityscapes:

  1. Download the pretrained ImageNet .pkl file.
  2. Put the model under pretrained_models/ folder with some file name [FILENAME].
  3. In the corresponding experiments/cityscapes/seg_MDEQ_[SIZE].yaml (where SIZE is typically SMALL, LARGE or XL), set MODEL.PRETRAINED to "pretrained_models/[FILENAME]".
  4. Run the MDEQ segmentation training command (see the "Usage" section above):
python -m torch.distributed.launch --nproc_per_node=[N_GPUS] tools/seg_train.py --cfg experiments/cityscapes/seg_MDEQ_[SIZE].yaml

III. Example of how to use the pretrained Cityscapes model for inference:

  1. Download the pretrained Cityscapes .pkl file
  2. Put the model under pretrained_models/ folder with some file name [FILENAME].
  3. In the corresponding experiments/cityscapes/seg_MDEQ_[SIZE].yaml (where SIZE is typically SMALL, LARGE or XL), set TEST.MODEL_FILE to "pretrained_models/[FILENAME]".
  4. Run the MDEQ segmentation testing command (see the "Usage" section above):
python tools/seg_test.py --cfg experiments/cityscapes/seg_MDEQ_[SIZE].yaml

Tips:

Acknowledgement

Some utilization code (e.g., model summary and yaml processing) of this repo were modified from the HRNet repo and the DEQ repo.