A PyTorch implementation of Mugs proposed by our paper "Mugs: A Multi-Granular Self-Supervised Learning Framework".

Related tags

Deep Learningmugs
Overview

Mugs: A Multi-Granular Self-Supervised Learning Framework

This is a PyTorch implementation of Mugs proposed by our paper "Mugs: A Multi-Granular Self-Supervised Learning Framework". arXiv

PWC

Overall framework of Mugs.

Fig 1. Overall framework of Mugs. In (a), for each image, two random crops of one image are fed into backbones of student and teacher. Three granular supervisions: 1) instance discrimination supervision, 2) local-group discrimination supervision, and 3) group discrimination supervision, are adopted to learn multi-granular representation. In (b), local-group modules in student/teacher averages all patch tokens, and finds top-k neighbors from memory buffer to aggregate them with the average for obtaining a local-group feature.

Pretrained models on ImageNet-1K

You can choose to download only the weights of the pretrained backbone used for downstream tasks, or the full checkpoint which contains backbone and projection head weights for both student and teacher networks.

Table 1. KNN and linear probing performance with their corresponding hyper-parameters, logs and model weights.

arch params pretraining epochs k-nn linear download
ViT-S/16 21M 100 72.3% 76.4% backbone only full ckpt args logs eval logs
ViT-S/16 21M 300 74.8% 78.2% backbone only full ckpt args logs eval logs
ViT-S/16 21M 800 75.6% 78.9% backbone only full ckpt args logs eval logs
ViT-B/16 85M 400 78.0% 80.6% backbone only full ckpt args logs eval logs
ViT-L/16 307M 250 80.3% 82.1% backbone only full ckpt args logs eval logs
Comparison of linear probing accuracy on ImageNet-1K.

Fig 2. Comparison of linear probing accuracy on ImageNet-1K.

Pretraining Settings

Environment

For reproducing, please install PyTorch and download the ImageNet dataset. This codebase has been developed with python version 3.8, PyTorch version 1.7.1, CUDA 11.0 and torchvision 0.8.2. For the full environment, please refer to our Dockerfile file.

ViT pretraining 🍺

To pretraining each model, please find the exact hyper-parameter settings at the args column of Table 1. For training log and linear probing log, please refer to the log and eval logs column of Table 1.

ViT-Small pretraining:

To run ViT-small for 100 epochs, we use two nodes of total 8 A100 GPUs (total 512 minibatch size) by using following command:

python -m torch.distributed.launch --nproc_per_node=8 main.py --data_path DATASET_ROOT --output_dir OUTPUT_ROOT --arch vit_small 
--group_teacher_temp 0.04 --group_warmup_teacher_temp_epochs 0 --weight_decay_end 0.2 --norm_last_layer false --epochs 100

To run ViT-small for 300 epochs, we use two nodes of total 16 A100 GPUs (total 1024 minibatch size) by using following command:

python -m torch.distributed.launch --nproc_per_node=16 main.py --data_path DATASET_ROOT --output_dir OUTPUT_ROOT --arch vit_small 
--group_teacher_temp 0.07 --group_warmup_teacher_temp_epochs 30 --weight_decay_end 0.1 --norm_last_layer false --epochs 300

To run ViT-small for 800 epochs, we use two nodes of total 16 A100 GPUs (total 1024 minibatch size) by using following command:

python -m torch.distributed.launch --nproc_per_node=16 main.py --data_path DATASET_ROOT --output_dir OUTPUT_ROOT --arch vit_small 
--group_teacher_temp 0.07 --group_warmup_teacher_temp_epochs 30 --weight_decay_end 0.1 --norm_last_layer false --epochs 800

ViT-Base pretraining:

To run ViT-base for 400 epochs, we use two nodes of total 24 A100 GPUs (total 1024 minibatch size) by using following command:

python -m torch.distributed.launch --nproc_per_node=24 main.py --data_path DATASET_ROOT --output_dir OUTPUT_ROOT --arch vit_base 
--group_teacher_temp 0.07 --group_warmup_teacher_temp_epochs 50 --min_lr 2e-06 --weight_decay_end 0.1 --freeze_last_layer 3 --norm_last_layer 
false --epochs 400

ViT-Large pretraining:

To run ViT-large for 250 epochs, we use two nodes of total 40 A100 GPUs (total 640 minibatch size) by using following command:

python -m torch.distributed.launch --nproc_per_node=40 main.py --data_path DATASET_ROOT --output_dir OUTPUT_ROOT --arch vit_large 
--lr 0.0015 --min_lr 1.5e-4 --group_teacher_temp 0.07 --group_warmup_teacher_temp_epochs 50 --weight_decay 0.025 
--weight_decay_end 0.08 --norm_last_layer true --drop_path_rate 0.3 --freeze_last_layer 3 --epochs 250

Evaluation

We are cleaning up the evalutation code and will release them when they are ready.

Self-attention visualization

Here we provide the self-attention map of the [CLS] token on the heads of the last layer

Self-attention from a ViT-Base/16 trained with Mugs

Fig 3. Self-attention from a ViT-Base/16 trained with Mugs.

T-SNE visualization

Here we provide the T-SNE visualization of the learned feature by ViT-B/16. We show the fish classes in ImageNet-1K, i.e., the first six classes, including tench, goldfish, white shark, tiger shark, hammerhead, electric ray. See more examples in Appendix.

T-SNE visualization of the learned feature by ViT-B/16.

Fig 4. T-SNE visualization of the learned feature by ViT-B/16.

License

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

Citation

If you find this repository useful, please consider giving a star and citation 🍺 :

@inproceedings{mugs2022SSL,
  title={Mugs: A Multi-Granular Self-Supervised Learning Framework},
  author={Pan Zhou and Yichen Zhou and Chenyang Si and Weihao Yu and Teck Khim Ng and Shuicheng Yan},
  booktitle={arXiv preprint arXiv:2203.14415},
  year={2022}
}
Owner
Sea AI Lab
Sea AI Lab
Human4D Dataset tools for processing and visualization

HUMAN4D: A Human-Centric Multimodal Dataset for Motions & Immersive Media HUMAN4D constitutes a large and multimodal 4D dataset that contains a variet

tofis 15 Nov 09, 2022
Ludwig is a toolbox that allows to train and evaluate deep learning models without the need to write code.

Translated in 🇰🇷 Korean/ Ludwig is a toolbox that allows users to train and test deep learning models without the need to write code. It is built on

Ludwig 8.7k Dec 31, 2022
An LSTM for time-series classification

Update 10-April-2017 And now it works with Python3 and Tensorflow 1.1.0 Update 02-Jan-2017 I updated this repo. Now it works with Tensorflow 0.12. In

Rob Romijnders 391 Dec 27, 2022
Security evaluation module with onnx, pytorch, and SecML.

🚀 🐼 🔥 PandaVision Integrate and automate security evaluations with onnx, pytorch, and SecML! Installation Starting the server without Docker If you

Maura Pintor 11 Apr 12, 2022
Human annotated noisy labels for CIFAR-10 and CIFAR-100.

Dataloader for CIFAR-N CIFAR-10N noise_label = torch.load('./data/CIFAR-10_human.pt') clean_label = noise_label['clean_label'] worst_label = noise_lab

<a href=[email protected]"> 117 Nov 30, 2022
InsTrim: Lightweight Instrumentation for Coverage-guided Fuzzing

InsTrim The paper: InsTrim: Lightweight Instrumentation for Coverage-guided Fuzzing Build Prerequisite llvm-8.0-dev clang-8.0 cmake = 3.2 Make git cl

75 Dec 23, 2022
Code for the paper Language as a Cognitive Tool to Imagine Goals in Curiosity Driven Exploration

IMAGINE: Language as a Cognitive Tool to Imagine Goals in Curiosity Driven Exploration This repo contains the code base of the paper Language as a Cog

Flowers Team 26 Dec 22, 2022
Visual odometry package based on hardware-accelerated NVIDIA Elbrus library with world class quality and performance.

Isaac ROS Visual Odometry This repository provides a ROS2 package that estimates stereo visual inertial odometry using the Isaac Elbrus GPU-accelerate

NVIDIA Isaac ROS 343 Jan 03, 2023
An implementation of Equivariant e2 convolutional kernals into a convolutional self attention network, applied to radio astronomy data.

EquivariantSelfAttention An implementation of Equivariant e2 convolutional kernals into a convolutional self attention network, applied to radio astro

2 Nov 09, 2021
QICK: Quantum Instrumentation Control Kit

QICK: Quantum Instrumentation Control Kit The QICK is a kit of firmware and software to use the Xilinx RFSoC to control quantum systems. It consists o

81 Dec 15, 2022
Animatable Neural Radiance Fields for Modeling Dynamic Human Bodies

To make the comparison with Animatable NeRF easier on the Human3.6M dataset, we save the quantitative results at here, which also contains the results of other methods, including Neural Body, D-NeRF,

ZJU3DV 359 Jan 08, 2023
Research code for CVPR 2021 paper "End-to-End Human Pose and Mesh Reconstruction with Transformers"

MeshTransformer ✨ This is our research code of End-to-End Human Pose and Mesh Reconstruction with Transformers. MEsh TRansfOrmer is a simple yet effec

Microsoft 473 Dec 31, 2022
Tooling for the Common Objects In 3D dataset.

CO3D: Common Objects In 3D This repository contains a set of tools for working with the Common Objects in 3D (CO3D) dataset. Download the dataset The

Facebook Research 724 Jan 06, 2023
Codebase for testing whether hidden states of neural networks encode discrete structures.

structural-probes Codebase for testing whether hidden states of neural networks encode discrete structures. Based on the paper A Structural Probe for

John Hewitt 349 Dec 17, 2022
Deep Latent Force Models

Deep Latent Force Models This repository contains a PyTorch implementation of the deep latent force model (DLFM), presented in the paper, Compositiona

Tom McDonald 5 Oct 26, 2022
PyTorch implementation of SIFT descriptor

This is an differentiable pytorch implementation of SIFT patch descriptor. It is very slow for describing one patch, but quite fast for batch. It can

Dmytro Mishkin 150 Dec 24, 2022
Dados coletados e programas desenvolvidos no processo de iniciação científica

Iniciacao_cientifica_FAPESP_2020-14845-6 Dados coletados e programas desenvolvidos no processo de iniciação científica Os arquivos .py são os programa

1 Jan 10, 2022
PyTorch implementation for the Neuro-Symbolic Sudoku Solver leveraging the power of Neural Logic Machines (NLM)

Neuro-Symbolic Sudoku Solver PyTorch implementation for the Neuro-Symbolic Sudoku Solver leveraging the power of Neural Logic Machines (NLM). Please n

Ashutosh Hathidara 60 Dec 10, 2022
Memory Defense: More Robust Classificationvia a Memory-Masking Autoencoder

Memory Defense: More Robust Classificationvia a Memory-Masking Autoencoder Authors: - Eashan Adhikarla - Dan Luo - Dr. Brian D. Davison Abstract Many

Eashan Adhikarla 4 Dec 25, 2022
Dewarping Document Image By Displacement Flow Estimation with Fully Convolutional Network.

Dewarping Document Image By Displacement Flow Estimation with Fully Convolutional Network

111 Dec 27, 2022