Face uncertainty quantification or estimation using PyTorch.

Overview

Face-uncertainty-pytorch

This is a demo code of face uncertainty quantification or estimation using PyTorch. The uncertainty of face recognition is affected by the ability of the recognition model (model uncertainty) and the quality of the input image (data uncertainty).

Model Uncertainty:

  • MC-Dropout

Data Uncertainty:

Usage

Preprocessing

Download the MS-Celeb-1M dataset from 1 or 2:

  1. insightface, https://github.com/deepinsight/insightface/wiki/Dataset-Zoo
  2. face.evoLVe.PyTorch, https://github.com/ZhaoJ9014/face.evoLVe.PyTorch#Data-Zoo)

Decode it using the code: https://github.com/deepinsight/insightface/blob/master/recognition/common/rec2image.py

Training

  1. Download the base model from https://github.com/deepinsight/insightface/tree/master/recognition/arcface_torch

  2. Modify the configuration files under config/ folder.

  3. Start the training:

    python network.py --config_file config/config_ir50_idq_loss_glint360k.py
    Start Training
    name: glint_ir50_idq
    num_epochs: 12
    epoch_size: 1000
    batch_size: 80
    num_c_in_batch 10 num_img_each_c 8.0
    IDQ_loss soft 16 0.45
    2022-01-12 23:37:48 [0-100] | loss 0.535 lr0.01 cos 0.55 1.00 0.18 pconf 0.77 1.00 0.15 t_soft 0.69 1.00 0.01 uloss 0.535 mem 3.1 G
    2022-01-12 23:38:12 [0-200] | loss 0.464 lr0.01 cos 0.58 0.93 0.08 pconf 0.75 1.00 0.05 t_soft 0.76 1.00 0.00 uloss 0.464 mem 3.1 G
    2022-01-12 23:38:37 [0-300] | loss 0.533 lr0.01 cos 0.52 1.00 0.04 pconf 0.78 0.99 0.25 t_soft 0.63 1.00 0.00 uloss 0.533 mem 3.1 G
    2022-01-12 23:39:02 [0-400] | loss 0.511 lr0.01 cos 0.52 0.99 0.09 pconf 0.77 0.99 0.16 t_soft 0.61 1.00 0.00 uloss 0.511 mem 3.1 G
    2022-01-12 23:39:27 [0-500] | loss 0.554 lr0.01 cos 0.48 0.97 0.05 pconf 0.77 0.99 0.18 t_soft 0.56 1.00 0.00 uloss 0.554 mem 3.1 G
    2022-01-12 23:39:52 [0-600] | loss 0.462 lr0.01 cos 0.55 0.95 0.19 pconf 0.78 0.99 0.23 t_soft 0.70 1.00 0.01 uloss 0.462 mem 3.1 G
    2022-01-12 23:40:17 [0-700] | loss 0.408 lr0.01 cos 0.55 0.96 0.07 pconf 0.78 0.99 0.07 t_soft 0.70 1.00 0.00 uloss 0.408 mem 3.1 G
    2022-01-12 23:40:42 [0-800] | loss 0.532 lr0.01 cos 0.51 0.99 0.03 pconf 0.80 0.99 0.25 t_soft 0.63 1.00 0.00 uloss 0.532 mem 3.1 G
    2022-01-12 23:41:06 [0-900] | loss 0.563 lr0.01 cos 0.54 1.00 0.03 pconf 0.80 0.99 0.13 t_soft 0.66 1.00 0.00 uloss 0.563 mem 3.1 G
    2022-01-12 23:41:27 [0-1000] | loss 0.570 lr0.01 cos 0.50 0.86 0.11 pconf 0.78 0.99 0.16 t_soft 0.61 1.00 0.00 uloss 0.570 mem 3.1 G
    ---cfp_fp
    sigma_sq [0.00263163 0.01750576 0.04416942 0.10698225 0.23958328 0.46090251
     0.92462665] percentile [0, 10, 30, 50, 70, 90, 100]
    reject_factor 0.0000 risk_threshold 0.924627 keep_idxes 7000 / 7000 Cosine score eer 0.012571 fmr100 0.012571 fmr1000 0.018286
    reject_factor 0.0500 risk_threshold 0.650710 keep_idxes 6655 / 7000 Cosine score eer 0.004357 fmr100 0.003900 fmr1000 0.006601
    reject_factor 0.1000 risk_threshold 0.556291 keep_idxes 6300 / 7000 Cosine score eer 0.003968 fmr100 0.003791 fmr1000 0.006003
    reject_factor 0.1500 risk_threshold 0.509630 keep_idxes 5951 / 7000 Cosine score eer 0.003864 fmr100 0.004013 fmr1000 0.005351
    reject_factor 0.2000 risk_threshold 0.459032 keep_idxes 5600 / 7000 Cosine score eer 0.003392 fmr100 0.003540 fmr1000 0.004248
    reject_factor 0.2500 risk_threshold 0.421400 keep_idxes 5251 / 7000 Cosine score eer 0.003236 fmr100 0.003407 fmr1000 0.003785
    reject_factor 0.3000 risk_threshold 0.389943 keep_idxes 4903 / 7000 Cosine score eer 0.002651 fmr100 0.002436 fmr1000 0.002842
    reject_factor mean --------------------------------------------- Cosine score fmr1000 0.002684
    AUERC: 0.0026
    AUERC30: 0.0017
    AUC: 0.0024
    AUC30: 0.0015
    

Testing

We use lfw.bin, cfp_fp.bin, etc. from ms1m-retinaface-t1 as the test dataset.

python evaluation/verification_risk_fnmr.py

MC-Dropout

python mc_dropout/verification_risk_mcdropout_fnmr.py
Owner
Kaen
Kaen
Godot RL Agents is a fully Open Source packages that allows video game creators

Godot RL Agents The Godot RL Agents is a fully Open Source packages that allows video game creators, AI researchers and hobbiest the opportunity to le

Edward Beeching 326 Dec 30, 2022
Generative Models as a Data Source for Multiview Representation Learning

GenRep Project Page | Paper Generative Models as a Data Source for Multiview Representation Learning Ali Jahanian, Xavier Puig, Yonglong Tian, Phillip

Ali 81 Dec 03, 2022
Predictive Modeling on Electronic Health Records(EHR) using Pytorch

Predictive Modeling on Electronic Health Records(EHR) using Pytorch Overview Although there are plenty of repos on vision and NLP models, there are ve

81 Jan 01, 2023
🗺 General purpose U-Network implemented in Keras for image segmentation

TF-Unet General purpose U-Network implemented in Keras for image segmentation Getting started • Training • Evaluation Getting started Looking for Jupy

Or Fleisher 2 Aug 31, 2022
Code for the paper "Implicit Representations of Meaning in Neural Language Models"

Implicit Representations of Meaning in Neural Language Models Preliminaries Create and set up a conda environment as follows: conda create -n state-pr

Belinda Li 39 Nov 03, 2022
VOGUE: Try-On by StyleGAN Interpolation Optimization

VOGUE is a StyleGAN interpolation optimization algorithm for photo-realistic try-on. Top: shirt try-on automatically synthesized by our method in two different examples.

Wei ZHANG 66 Dec 09, 2022
Supervised 3D Pre-training on Large-scale 2D Natural Image Datasets for 3D Medical Image Analysis

Introduction This is an implementation of our paper Supervised 3D Pre-training on Large-scale 2D Natural Image Datasets for 3D Medical Image Analysis.

24 Dec 06, 2022
Official PyTorch Implementation of "Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs". NeurIPS 2020.

Self-supervised Auxiliary Learning with Meta-paths for Heterogeneous Graphs This repository is the implementation of SELAR. Dasol Hwang* , Jinyoung Pa

MLV Lab (Machine Learning and Vision Lab at Korea University) 48 Nov 09, 2022
Config files for my GitHub profile.

Canalyst Candas Data Science Library Name Canalyst Candas Description Built by a former PM / analyst to give anyone with a little bit of Python knowle

Canalyst Candas 13 Jun 24, 2022
In this project we predict the forest cover type using the cartographic variables in the training/test datasets.

Kaggle Competition: Forest Cover Type Prediction In this project we predict the forest cover type (the predominant kind of tree cover) using the carto

Marianne Joy Leano 1 Mar 15, 2022
The official code repo of "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound Classification and Detection"

Hierarchical Token Semantic Audio Transformer Introduction The Code Repository for "HTS-AT: A Hierarchical Token-Semantic Audio Transformer for Sound

Knut(Ke) Chen 134 Jan 01, 2023
Age Progression/Regression by Conditional Adversarial Autoencoder

Age Progression/Regression by Conditional Adversarial Autoencoder (CAAE) TensorFlow implementation of the algorithm in the paper Age Progression/Regre

Zhifei Zhang 603 Dec 22, 2022
PiRapGenerator - Make anyone rap the digits of pi

PiRapGenerator Make anyone rap the digits of pi (sample files are of Ted Nivison

7 Oct 02, 2022
CodeContests is a competitive programming dataset for machine-learning

CodeContests CodeContests is a competitive programming dataset for machine-learning. This dataset was used when training AlphaCode. It consists of pro

DeepMind 1.6k Jan 08, 2023
免费获取http代理并生成proxifier配置文件

freeproxy 免费获取http代理并生成proxifier配置文件 公众号:台下言书 工具说明:https://mp.weixin.qq.com/s?__biz=MzIyNDkwNjQ5Ng==&mid=2247484425&idx=1&sn=56ccbe130822aa35038095317

说书人 32 Mar 25, 2022
Official PyTorch Implementation of Learning Architectures for Binary Networks

Learning Architectures for Binary Networks An Pytorch Implementation of the paper Learning Architectures for Binary Networks (BNAS) (ECCV 2020) If you

Computer Vision Lab. @ GIST 25 Jun 09, 2022
Official PyTorch implementation of the paper: DeepSIM: Image Shape Manipulation from a Single Augmented Training Sample

DeepSIM: Image Shape Manipulation from a Single Augmented Training Sample (ICCV 2021 Oral) Project | Paper Official PyTorch implementation of the pape

Eliahu Horwitz 393 Dec 22, 2022
Perception-aware multi-sensor fusion for 3D LiDAR semantic segmentation (ICCV 2021)

Perception-Aware Multi-Sensor Fusion for 3D LiDAR Semantic Segmentation (ICCV 2021) [中文|EN] 概述 本工作主要探索一种高效的多传感器(激光雷达和摄像头)融合点云语义分割方法。现有的多传感器融合方法主要将点云投影

ICE 126 Dec 30, 2022
ICLR 2021 i-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning

Introduction PyTorch code for the ICLR 2021 paper [i-Mix: A Domain-Agnostic Strategy for Contrastive Representation Learning]. @inproceedings{lee2021i

Kibok Lee 68 Nov 27, 2022
My solutions for Stanford University course CS224W: Machine Learning with Graphs Fall 2021 colabs (GNN, GAT, GraphSAGE, GCN)

machine-learning-with-graphs My solutions for Stanford University course CS224W: Machine Learning with Graphs Fall 2021 colabs Course materials can be

Marko Njegomir 7 Dec 14, 2022