Asymmetric metric learning for knowledge transfer

Related tags

Deep Learningaml
Overview

Asymmetric metric learning

This is the official code that enables the reproduction of the results from our paper:

Asymmetric metric learning for knowledge transfer, Budnik M., Avrithis Y. [arXiv]

Content

This repository provides the means to train and test all the models presented in the paper. This includes:

  1. Code to train the models with and without the teacher (asymmetric and symmetric).
  2. Code to do symmetric and asymmetric testing on rOxford and rParis datasets.
  3. Best pre-trainend models (including whitening).

Dependencies

  1. Python3 (tested on version 3.6)
  2. Numpy 1.19
  3. PyTorch (tested on version 1.4.0)
  4. Datasets and base models will be downloaded automatically.

Training and testing the networks

To train a model use the following script:

python main.py [-h] [--training-dataset DATASET] [--directory EXPORT_DIR] [--no-val]
                  [--test-datasets DATASETS] [--test-whiten DATASET]
                  [--val-freq N] [--save-freq N] [--arch ARCH] [--pool POOL]
                  [--local-whitening] [--regional] [--whitening]
                  [--not-pretrained] [--loss LOSS] [--loss-margin LM] 
                  [--mode MODE] [--teacher TEACHER] [--sym]
                  [--feat-path FEAT] [--feat-val-path FEATVAL]
                  [--image-size N] [--neg-num N] [--query-size N]
                  [--pool-size N] [--gpu-id N] [--workers N] [--epochs N]
                  [--batch-size N] [--optimizer OPTIMIZER] [--lr LR]
                  [--momentum M] [--weight-decay W] [--print-freq N]
                  [--resume FILENAME] [--comment COMMENT] 
                  

Most parameters are the same as in CNN Image Retrieval in PyTorch. Here, we describe parameters added or modified in this work, namely:
--arch - architecture of the model to be trained, in our case the student.
--mode - is the training mode, which determines how the dataset is handled, e.g. are the tuples constructed randomly or with mining; which examples are coming from the teacher vs student, etc. So for example while the --loss is set to 'contrastive', 'ts' enables standard student-teacher training (includes mining), 'ts_self' trains using the Contr+ approach, 'reg' uses the regression. When using 'rand' or 'reg' no mining is used. With 'std' it follows the original training protocol from here (the teacher model is not used).
--teacher - the model of the teacher(vgg16 or resnet101), note that this param makes the last layer of the student match that of the teacher. Therefore, this can be used even in a standard symmetric training.
--sym - a flag that indicates if the training should be symmetric or asymmetric.
--feat-path and --feat-val-path - a path to the extracted teacher features used to train the student. The features can be extracted using the extract_features.py script.

To perform a symmetric test of the model that is already trained:

python test.py [-h] (--network-path NETWORK | --network-offtheshelf NETWORK)
               [--datasets DATASETS] [--image-size N] [--multiscale MULTISCALE] 
               [--whitening WHITENING] [--teacher TEACHER]

For the asymmetric testing:

python test.py [-h] (--network-path NETWORK | --network-offtheshelf NETWORK)
               [--datasets DATASETS] [--image-size N] [--multiscale MULTISCALE] 
               [--whitening WHITENING] [--teacher TEACHER] [--asym]

Examples:

Perform a symmetric test with a pre-trained model:

python test.py -npath  mobilenet-v2-gem-contr-vgg16 -d 'roxford5k,rparis6k' -ms '[1, 1/2**(1/2), 1/2]' -w retrieval-SfM-120k --teacher vgg16

For an asymmetric test:

python test.py -npath  mobilenet-v2-gem-contr-vgg16 -d 'roxford5k,rparis6k' -ms '[1, 1/2**(1/2), 1/2]' -w retrieval-SfM-120k --teacher vgg16 --asym

If you are interested in just the trained models, you can find the links to them in the test.py file.

Acknowledgements

This code is adapted and modified based on the amazing repository by F. Radenović called CNN Image Retrieval in PyTorch: Training and evaluating CNNs for Image Retrieval in PyTorch

IJCAI2020 & IJCV 2020 :city_sunrise: Unsupervised Scene Adaptation with Memory Regularization in vivo

Seg_Uncertainty In this repo, we provide the code for the two papers, i.e., MRNet:Unsupervised Scene Adaptation with Memory Regularization in vivo, IJ

Zhedong Zheng 348 Jan 05, 2023
Orthogonal Over-Parameterized Training

The inductive bias of a neural network is largely determined by the architecture and the training algorithm. To achieve good generalization, how to effectively train a neural network is of great impo

Weiyang Liu 11 Apr 18, 2022
More Photos are All You Need: Semi-Supervised Learning for Fine-Grained Sketch Based Image Retrieval

More Photos are All You Need: Semi-Supervised Learning for Fine-Grained Sketch Based Image Retrieval, CVPR 2021. Ayan Kumar Bhunia, Pinaki nath Chowdh

Ayan Kumar Bhunia 22 Aug 27, 2022
Trax — Deep Learning with Clear Code and Speed

Trax — Deep Learning with Clear Code and Speed Trax is an end-to-end library for deep learning that focuses on clear code and speed. It is actively us

Google 7.3k Dec 26, 2022
[3DV 2020] PeeledHuman: Robust Shape Representation for Textured 3D Human Body Reconstruction

PeeledHuman: Robust Shape Representation for Textured 3D Human Body Reconstruction International Conference on 3D Vision, 2020 Sai Sagar Jinka1, Rohan

Rohan Chacko 39 Oct 12, 2022
Compartmental epidemic model to assess undocumented infections: applications to SARS-CoV-2 epidemics in Brazil - Datasets and Codes

Compartmental epidemic model to assess undocumented infections: applications to SARS-CoV-2 epidemics in Brazil - Datasets and Codes The codes for simu

1 Jan 12, 2022
Automatic deep learning for image classification.

AutoDL AutoDL automates machine learning tasks enabling you to easily achieve strong predictive performance in your applications. With just a few line

wenqi 2 Oct 12, 2022
Open-Set Recognition: A Good Closed-Set Classifier is All You Need

Open-Set Recognition: A Good Closed-Set Classifier is All You Need Code for our paper: "Open-Set Recognition: A Good Closed-Set Classifier is All You

194 Jan 03, 2023
Generalized Random Forests

generalized random forests A pluggable package for forest-based statistical estimation and inference. GRF currently provides non-parametric methods fo

GRF Labs 781 Dec 25, 2022
Implement slightly different caffe-segnet in tensorflow

Tensorflow-SegNet Implement slightly different (see below for detail) SegNet in tensorflow, successfully trained segnet-basic in CamVid dataset. Due t

Tseng Kuan Lun 364 Oct 27, 2022
PyTorchMemTracer - Depict GPU memory footprint during DNN training of PyTorch

A Memory Tracer For PyTorch OOM is a nightmare for PyTorch users. However, most

Jiarui Fang 9 Nov 14, 2022
Some bravo or inspiring research works on the topic of curriculum learning.

Towards Scalable Unpaired Virtual Try-On via Patch-Routed Spatially-Adaptive GAN Official code for NeurIPS 2021 paper "Towards Scalable Unpaired Virtu

131 Jan 07, 2023
LegoDNN: a block-grained scaling tool for mobile vision systems

Table of contents 1 Introduction 1.1 Major features 1.2 Architecture 2 Code and Installation 2.1 Code 2.2 Installation 3 Repository of DNNs in vision

41 Dec 24, 2022
Tool which allow you to detect and translate text.

Text detection and recognition This repository contains tool which allow to detect region with text and translate it one by one. Description Two pretr

Damian Panek 176 Nov 28, 2022
Algorithmic encoding of protected characteristics and its implications on disparities across subgroups

Algorithmic encoding of protected characteristics and its implications on disparities across subgroups This repository contains the code for the paper

Team MIRA - BioMedIA 15 Oct 24, 2022
A chemical analysis of lipophilicities & molecule drawings including ML

A chemical analysis of lipophilicity & molecule drawings including a bit of ML analysis. This is a simple project that includes two Jupyter files (one

Aurimas A. Nausėdas 7 Nov 22, 2022
A minimalist implementation of score-based diffusion model

sdeflow-light This is a minimalist codebase for training score-based diffusion models (supporting MNIST and CIFAR-10) used in the following paper "A V

Chin-Wei Huang 89 Dec 20, 2022
The official implementation of ICCV paper "Box-Aware Feature Enhancement for Single Object Tracking on Point Clouds".

Box-Aware Tracker (BAT) Pytorch-Lightning implementation of the Box-Aware Tracker. Box-Aware Feature Enhancement for Single Object Tracking on Point C

Kangel Zenn 5 Mar 26, 2022
This is a work in progress reimplementation of Instant Neural Graphics Primitives

Neural Hash Encoding This is a work in progress reimplementation of Instant Neural Graphics Primitives Currently this can train an implicit representa

Penn 79 Sep 01, 2022
Self-Supervised Pre-Training for Transformer-Based Person Re-Identification

Self-Supervised Pre-Training for Transformer-Based Person Re-Identification [pdf] The official repository for Self-Supervised Pre-Training for Transfo

Hao Luo 116 Jan 04, 2023