Code Monkey home page Code Monkey logo

fastconvmae's Introduction

🚀Fast ConvMAE🚀

Fast ConvMAE: Fast Pretraining of ConvMAE

This repo is the faster implementation of ConvMAE: Masked Convolution Meets Masked Autoencoders

Updates

17/June/2022

Released the pre-training codes for ImageNet-1K.

Introduction

Fast ConvMAE framework is a superiorly fast masked modeling scheme via complementary masking and mixture of reconstrunctors based on the ConvMAE.

tenser

Pretrain on ImageNet-1K

The following table provides pretrained checkpoints and logs used in the paper.

Fast ConvMAE-Base
50epoch pretrained checkpoints N/A
logs N/A

Main Results on COCO & ImageNet-1K

Models Masking Tokenizer Backbone PT Epochs PT Hours COCO FT Epochs $AP^{Box}$ $AP^{Mask}$ ImageNet Finetune Epochs Finetune acc@1(%) ADE 20K mIoU
ConvMAE 25 % RGB ConvViT-B 200 512 25 50.8 45.4 100 84.4 48.5
ConvMAE 25 % RGB ConvViT-B 1600 4000 25 53.2 47.1 100 85.0 51.7
MAE 25 % RGB ViT-B 1600 2069 100 50.3 44.9 100 83.6 48.1
SimMIM 100 % RGB Swin-B 800 1609 36 50.4 44.4 100 84.0 -
GreenMIM 25 % RGB Swin-B 800 887 36 50.0 44.1 100 85.1 -
ConvMAE 100 % RGB ConvViT-B 50 266 25 51.0 45.4 100 84.4 48.3
ConvMAE 100 % C+T ConvViT-B 50 333 25 52.8 46.9 100 85.0 52.7
ConvMAE 100 % C+T ConvViT-B 100 666 25 53.3 47.3 100 85.2 52.8
ConvMAE 100 % C+T ConvViT-L 200 N/A 25 N/A N/A 50 86.7 54.5

Visualizations

NOTE: Grey patches are masked and colored ones are kept.

tenser

Getting Started

Prerequisites

  • Linux
  • Python 3.7+
  • CUDA 10.2+
  • GCC 5+

Training and evaluation

Acknowledgement

The pretraining and finetuning of our project are based on DeiT, MAE, and ConvMAE. Thanks for their wonderful work.

License

FastConvMAE is released under the MIT License.

visitors

fastconvmae's People

Contributors

alpha-vl avatar gaopengpjlab avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.