ML-Decoder: Scalable and Versatile Classification Head

Overview

ML-Decoder: Scalable and Versatile Classification Head

PWC
PWC
PWC


Paper

Official PyTorch Implementation

Tal Ridnik, Gilad Sharir, Avi Ben-Cohen, Emanuel Ben-Baruch, Asaf Noy
DAMO Academy, Alibaba Group

Abstract

In this paper, we introduce ML-Decoder, a new attention-based classification head. ML-Decoder predicts the existence of class labels via queries, and enables better utilization of spatial data compared to global average pooling. By redesigning the decoder architecture, and using a novel group-decoding scheme, ML-Decoder is highly efficient, and can scale well to thousands of classes. Compared to using a larger backbone, ML-Decoder consistently provides a better speed-accuracy trade-off. ML-Decoder is also versatile - it can be used as a drop-in replacement for various classification heads, and generalize to unseen classes when operated with word queries. Novel query augmentations further improve its generalization ability. Using ML-Decoder, we achieve state-of-the-art results on several classification tasks: on MS-COCO multi-label, we reach 91.4% mAP; on NUS-WIDE zero-shot, we reach 31.1% ZSL mAP; and on ImageNet single-label, we reach with vanilla ResNet50 backbone a new top score of 80.7%, without extra data or distillation.

ML-Decoder Implementation

ML-Decoder implementation is available here. It can be easily integrated into any backbone using this example code:

ml_decoder_head = MLDecoder(num_classes) # initilization

spatial_embeddings = self.backbone(input_image) # backbone generates spatial embeddings      
 
logits = ml_decoder_head(spatial_embeddings) # transfrom spatial embeddings to logits

Training Code

We will share a full reproduction code for the article results.

Multi-label Training Code


A reproduction code for MS-COCO multi-label:

python train.py  \
--data=/home/datasets/coco2014/ \
--model_name=tresnet_l \
--image_size=448

Single-label Training Code

Our single-label training code uses the excellent timm repo. Reproduction code is currently from a fork, we will work toward a full merge to the main repo.

git clone https://github.com/mrT23/pytorch-image-models.git

This is the code for A2 configuration training, with ML-Decoder (--use-ml-decoder-head=1):

python -u -m torch.distributed.launch --nproc_per_node=8 \
--nnodes=1 \
--node_rank=0 \
./train.py \
/data/imagenet/ \
--amp \
-b=256 \
--epochs=300 \
--drop-path=0.05 \
--opt=lamb \
--weight-decay=0.02 \
--sched='cosine' \
--lr=4e-3 \
--warmup-epochs=5 \
--model=resnet50 \
--aa=rand-m7-mstd0.5-inc1 \
--reprob=0.0 \
--remode='pixel' \
--mixup=0.1 \
--cutmix=1.0 \
--aug-repeats 3 \
--bce-target-thresh 0.2 \
--smoothing=0 \
--bce-loss \
--train-interpolation=bicubic \
--use-ml-decoder-head=1

ZSL Training Code

Reproduction code for ZSL is WIP.

Citation

@misc{ridnik2021mldecoder,
      title={ML-Decoder: Scalable and Versatile Classification Head}, 
      author={Tal Ridnik and Gilad Sharir and Avi Ben-Cohen and Emanuel Ben-Baruch and Asaf Noy},
      year={2021},
      eprint={2111.12933},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}
Spatially-Adaptive Pixelwise Networks for Fast Image Translation, CVPR 2021

Image Translation with ASAPNets Spatially-Adaptive Pixelwise Networks for Fast Image Translation, CVPR 2021 Webpage | Paper | Video Installation insta

Tamar Rott Shaham 100 Dec 28, 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
Python and Julia in harmony.

PythonCall & JuliaCall Bringing Python® and Julia together in seamless harmony: Call Python code from Julia and Julia code from Python via a symmetric

Christopher Rowley 414 Jan 07, 2023
An adaptive hierarchical energy management strategy for hybrid electric vehicles

An adaptive hierarchical energy management strategy This project contains the source code of an adaptive hierarchical EMS combining heuristic equivale

19 Dec 13, 2022
UMich 500-Level Mobile Robotics Course

MOBILE ROBOTICS: METHODS & ALGORITHMS - WINTER 2022 University of Michigan - NA 568/EECS 568/ROB 530 For slides, lecture notes, and example codes, see

393 Dec 29, 2022
Code for unmixing audio signals in four different stems "drums, bass, vocals, others". The code is adapted from "Jukebox: A Generative Model for Music"

Status: Archive (code is provided as-is, no updates expected) Disclaimer This code is a based on "Jukebox: A Generative Model for Music" Paper We adju

Wadhah Zai El Amri 24 Dec 29, 2022
Reverse engineering recurrent neural networks with Jacobian switching linear dynamical systems

Reverse engineering recurrent neural networks with Jacobian switching linear dynamical systems This repository is the official implementation of Rever

6 Aug 25, 2022
System Design course at HSE (2021)

System Design course at HSE (2021) Wiki-страница курса Структура репозитория: slides - директория с презентациями с занятий tasks - материалы для выпо

22 Dec 25, 2022
Retinal vessel segmentation based on GT-UNet

Retinal vessel segmentation based on GT-UNet Introduction This project is a retinal blood vessel segmentation code based on UNet-like Group Transforme

Kent0n 27 Dec 18, 2022
NeuralCompression is a Python repository dedicated to research of neural networks that compress data

NeuralCompression is a Python repository dedicated to research of neural networks that compress data. The repository includes tools such as JAX-based entropy coders, image compression models, video c

Facebook Research 297 Jan 06, 2023
PyTorch implemention of ICCV'21 paper SGPA: Structure-Guided Prior Adaptation for Category-Level 6D Object Pose Estimation

SGPA: Structure-Guided Prior Adaptation for Category-Level 6D Object Pose Estimation This is the PyTorch implemention of ICCV'21 paper SGPA: Structure

Chen Kai 24 Dec 05, 2022
YOLOX + ROS(1, 2) object detection package

YOLOX + ROS(1, 2) object detection package

Ar-Ray 158 Dec 21, 2022
Python code to generate art with Generative Adversarial Network

GAN_Canvas_Maker Generating Art using Generative Adversarial Network (GAN) Python code to generate art with Generative Adversarial Network: https://to

Jonny Banana 10 Aug 22, 2022
Image Deblurring using Generative Adversarial Networks

DeblurGAN arXiv Paper Version Pytorch implementation of the paper DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks. Our netwo

Orest Kupyn 2.2k Jan 01, 2023
Code for 'Self-Guided and Cross-Guided Learning for Few-shot segmentation. (CVPR' 2021)'

SCL Introduction Code for 'Self-Guided and Cross-Guided Learning for Few-shot segmentation. (CVPR' 2021)' We evaluated our approach using two baseline

34 Oct 08, 2022
CR-Fill: Generative Image Inpainting with Auxiliary Contextual Reconstruction. ICCV 2021

crfill Usage | Web App | | Paper | Supplementary Material | More results | code for paper ``CR-Fill: Generative Image Inpainting with Auxiliary Contex

182 Dec 20, 2022
Code for the paper "A Study of Face Obfuscation in ImageNet"

A Study of Face Obfuscation in ImageNet Code for the paper: A Study of Face Obfuscation in ImageNet Kaiyu Yang, Jacqueline Yau, Li Fei-Fei, Jia Deng,

35 Oct 04, 2022
Code for the paper SphereRPN: Learning Spheres for High-Quality Region Proposals on 3D Point Clouds Object Detection, ICIP 2021.

SphereRPN Code for the paper SphereRPN: Learning Spheres for High-Quality Region Proposals on 3D Point Clouds Object Detection, ICIP 2021. Authors: Th

Thang Vu 15 Dec 02, 2022
MAGMA - a GPT-style multimodal model that can understand any combination of images and language

MAGMA -- Multimodal Augmentation of Generative Models through Adapter-based Finetuning Authors repo (alphabetical) Constantin (CoEich), Mayukh (Mayukh

Aleph Alpha GmbH 331 Jan 03, 2023
PAWS 🐾 Predicting View-Assignments with Support Samples

This repo provides a PyTorch implementation of PAWS (predicting view assignments with support samples), as described in the paper Semi-Supervised Learning of Visual Features by Non-Parametrically Pre

Facebook Research 437 Dec 23, 2022