Code for Efficient Visual Pretraining with Contrastive Detection

Related tags

Deep Learningdetcon
Overview

Code for DetCon

This repository contains code for the ICCV 2021 paper "Efficient Visual Pretraining with Contrastive Detection" by Olivier J. Hénaff, Skanda Koppula, Jean-Baptiste Alayrac, Aaron van den Oord, Oriol Vinyals, João Carreira.

This repository includes sample code to run pretraining with DetCon. In particular, we're providing a sample script for generating the Felzenzwalb segmentations for ImageNet images (using skimage) and a pre-training experiment setup (dataloader, augmentation pipeline, optimization config, and loss definition) that describes the DetCon-B(YOL) model described in the paper. The original code uses a large grid of TPUs and internal infrastructure for training, but we've extracted the key DetCon loss+experiment in this folder so that external groups can have a reference should they want to explore a similar approaches.

This repository builds heavily from the BYOL open source release, so speed-up tricks and features in that setup may likely translate to the code here.

Running this code

Running ./setup.sh will create and activate a virtualenv and install all necessary dependencies. To enter the environment after running setup.sh, run source /tmp/detcon_venv/bin/activate.

Running bash test.sh will run a single training step on a mock image/Felzenszwalb mask as a simple validation that all dependencies are set up correctly and the DetCon pre-training can run smoothly. On our 16-core machine, running on CPU, we find this takes around 3-4 minutes.

A TFRecord dataset containing each ImageNet image, label, and its corresponding Felzenszwalb segmentation/mask can then be generated using the generate_fh_masks Python script. You will first have to download two pieces of ImageNet metadata into the same directory as the script:

wget https://raw.githubusercontent.com/tensorflow/models/master/research/slim/datasets/imagenet_metadata.txt wget https://raw.githubusercontent.com/tensorflow/models/master/research/slim/datasets/imagenet_lsvrc_2015_synsets.txt

And to run the multi-threaded mask generation script:

python generate_fh_masks_for_imagenet.py -- \
--train_directory=imagenet-train \
--output_directory=imagenet-train-fh

This single-machine, multi-threaded version of the mask generation script takes 2-3 days on a 16-core CPU machine to complete CPU-based processing of the ImageNet training and validation set. The script assumes the same ImageNet directory structure as github.com/tensorflow/models/blob/master/research/slim/datasets/build_imagenet_data.py (more details in the link).

You can then run the main training loop and execute multiple DetCon-B training steps by running from the parent directory the command:

python -m detcon.main_loop \
  --dataset_directory='/tmp/imagenet-fh-train' \
  --pretrain_epochs=100`

Note that you will need to update the dataset_directory flag, to point to the generated Felzenzwalb/image TFRecord dataset previously generated. Additionally, to use accelerators, users will need to install the correct version of jaxlib with CUDA support.

Citing this work

If you use this code in your work, please consider referencing our work:

@article{henaff2021efficient,
  title={{Efficient Visual Pretraining with Contrastive Detection}},
  author={H{\'e}naff, Olivier J and Koppula, Skanda and Alayrac, Jean-Baptiste and Oord, Aaron van den and Vinyals, Oriol and Carreira, Jo{\~a}o},
  journal={International Conference on Computer Vision},
  year={2021}
}

Disclaimer

This is not an officially supported Google product.

Owner
DeepMind
DeepMind
MaRS - a recursive filtering framework that allows for truly modular multi-sensor integration

The Modular and Robust State-Estimation Framework, or short, MaRS, is a recursive filtering framework that allows for truly modular multi-sensor integration

Control of Networked Systems - University of Klagenfurt 143 Dec 29, 2022
Activity image-based video retrieval

Cross-modal-retrieval Our approach is focus on Activity Image-to-Video Retrieval (AIVR) task. The compared methods are state-of-the-art single modalit

BCMI 75 Oct 21, 2021
Tensorflow implementation of Fully Convolutional Networks for Semantic Segmentation

FCN.tensorflow Tensorflow implementation of Fully Convolutional Networks for Semantic Segmentation (FCNs). The implementation is largely based on the

Sarath Shekkizhar 1.3k Dec 25, 2022
Official code for 'Robust Siamese Object Tracking for Unmanned Aerial Manipulator' and offical introduction to UAMT100 benchmark

SiamSA: Robust Siamese Object Tracking for Unmanned Aerial Manipulator Demo video 📹 Our video on Youtube and bilibili demonstrates the evaluation of

Intelligent Vision for Robotics in Complex Environment 12 Dec 18, 2022
Decentralized Reinforcment Learning: Global Decision-Making via Local Economic Transactions (ICML 2020)

Decentralized Reinforcement Learning This is the code complementing the paper Decentralized Reinforcment Learning: Global Decision-Making via Local Ec

40 Oct 30, 2022
UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning

UniMoCo: Unsupervised, Semi-Supervised and Full-Supervised Visual Representation Learning This is the official PyTorch implementation for UniMoCo pape

dddzg 49 Jan 02, 2023
Automated Melanoma Recognition in Dermoscopy Images via Very Deep Residual Networks

Introduction This repository contains the modified caffe library and network architectures for our paper "Automated Melanoma Recognition in Dermoscopy

Lequan Yu 47 Nov 24, 2022
GeneDisco is a benchmark suite for evaluating active learning algorithms for experimental design in drug discovery.

GeneDisco is a benchmark suite for evaluating active learning algorithms for experimental design in drug discovery.

22 Dec 12, 2022
Some pre-commit hooks for OpenMMLab projects

pre-commit-hooks Some pre-commit hooks for OpenMMLab projects. Using pre-commit-hooks with pre-commit Add this to your .pre-commit-config.yaml - rep

OpenMMLab 16 Nov 29, 2022
Official repository for the NeurIPS 2021 paper Get Fooled for the Right Reason: Improving Adversarial Robustness through a Teacher-guided curriculum Learning Approach

Get Fooled for the Right Reason Official repository for the NeurIPS 2021 paper Get Fooled for the Right Reason: Improving Adversarial Robustness throu

Sowrya Gali 1 Apr 25, 2022
"Segmenter: Transformer for Semantic Segmentation" reproduced via mmsegmentation

Segmenter-based-on-OpenMMLab "Segmenter: Transformer for Semantic Segmentation, arxiv 2105.05633." reproduced via mmsegmentation. We reproduce Segment

EricKani 22 Feb 24, 2022
Official code repository for ICCV 2021 paper: Gravity-Aware Monocular 3D Human Object Reconstruction

GraviCap Official code repository for ICCV 2021 paper: Gravity-Aware Monocular 3D Human Object Reconstruction. Gravity-Aware Monocular 3D Human-Object

Rishabh Dabral 15 Dec 09, 2022
A tool for calculating distortion parameters in coordination complexes.

OctaDist Octahedral distortion calculator: A tool for calculating distortion parameters in coordination complexes. https://octadist.github.io/ Registe

OctaDist 12 Oct 04, 2022
Computer Vision and Pattern Recognition, NUS CS4243, 2022

CS4243_2022 Computer Vision and Pattern Recognition, NUS CS4243, 2022 Cloud Machine #1 : Google Colab (Free GPU) Follow this Notebook installation : h

Xavier Bresson 142 Dec 15, 2022
68 keypoint annotations for COFW test data

68 keypoint annotations for COFW test data This repository contains manually annotated 68 keypoints for COFW test data (original annotation of CFOW da

31 Dec 06, 2022
Propagate Yourself: Exploring Pixel-Level Consistency for Unsupervised Visual Representation Learning, CVPR 2021

Propagate Yourself: Exploring Pixel-Level Consistency for Unsupervised Visual Representation Learning By Zhenda Xie*, Yutong Lin*, Zheng Zhang, Yue Ca

Zhenda Xie 293 Dec 20, 2022
JFB: Jacobian-Free Backpropagation for Implicit Models

JFB: Jacobian-Free Backpropagation for Implicit Models

Typal Research 28 Dec 11, 2022
TensorLight - A high-level framework for TensorFlow

TensorLight is a high-level framework for TensorFlow-based machine intelligence applications. It reduces boilerplate code and enables advanced feature

Benjamin Kan 10 Jul 31, 2022
PEPit is a package enabling computer-assisted worst-case analyses of first-order optimization methods.

PEPit: Performance Estimation in Python This open source Python library provides a generic way to use PEP framework in Python. Performance estimation

Baptiste 53 Nov 16, 2022
Aquarius - Enabling Fast, Scalable, Data-Driven Virtual Network Functions

Aquarius Aquarius - Enabling Fast, Scalable, Data-Driven Virtual Network Functions NOTE: We are currently going through the open-source process requir

Zhiyuan YAO 0 Jun 02, 2022