Reliable probability face embeddings

Related tags

Deep LearningProbFace
Overview

ProbFace, arxiv

This is a demo code of training and testing [ProbFace] using Tensorflow. ProbFace is a reliable Probabilistic Face Embeddging (PFE) method. The representation of each face will be an Guassian distribution parametrized by (mu, sigma), where mu is the original embedding and sigma is the learned uncertainty. Experiments show that ProbFace could

  • improve the robustness of PFE.
  • simplify the calculation of the multal likelihood score (MLS).
  • improve the recognition performance on the risk-controlled scenarios.

Usage

Preprocessing

Download the MS-Celeb-1M dataset from insightface or face.evoLVe.PyTorch and decode it using this code

Training

  1. Download the base model ResFace64 and unzip the files under log/resface64.

  2. Modify the configuration files under configfig/ folder.

  3. Start the training:

    python train.py configfig/resface64_msarcface.py
    Start Training
    name: resface64
    # epochs: 12
    epoch_size: 1000
    batch_size: 128
    
    Saving variables...
    Saving metagraph...
    Saving variables...
    [1][1] time: 4.19 a 0.8130 att_neg 2.7123 att_pos 0.9874 atte 1.8354 lr 0.0100 mls 0.6820 regu 0.1267 s_L2 0.0025 s_max 0.4467 s_min 0.2813
    [1][101] time: 37.72 a 0.8273 att_neg 2.9455 att_pos 1.0839 atte 1.8704 lr 0.0100 mls 0.6946 regu 0.1256 s_L2 0.0053 s_max 0.4935 s_min 0.2476
    [1][201] time: 38.06 a 0.8533 att_neg 2.9560 att_pos 1.1092 atte 1.9117 lr 0.0100 mls 0.7208 regu 0.1243 s_L2 0.0063 s_max 0.5041 s_min 0.2505
    [1][301] time: 38.82 a 0.7510 att_neg 2.9985 att_pos 1.0223 atte 1.7441 lr 0.0100 mls 0.6209 regu 0.1231 s_L2 0.0053 s_max 0.4552 s_min 0.2251
    [1][401] time: 37.95 a 0.8122 att_neg 2.9846 att_pos 1.0803 atte 1.8501 lr 0.0100 mls 0.6814 regu 0.1219 s_L2 0.0070 s_max 0.4964 s_min 0.2321
    [1][501] time: 38.42 a 0.7307 att_neg 3.0087 att_pos 1.0050 atte 1.8465 lr 0.0100 mls 0.6005 regu 0.1207 s_L2 0.0076 s_max 0.5249 s_min 0.2181
    [1][601] time: 37.69 a 0.7827 att_neg 3.0395 att_pos 1.0703 atte 1.8236 lr 0.0100 mls 0.6552 regu 0.1195 s_L2 0.0062 s_max 0.4952 s_min 0.2211
    [1][701] time: 37.36 a 0.7410 att_neg 2.9971 att_pos 1.0180 atte 1.8086 lr 0.0100 mls 0.6140 regu 0.1183 s_L2 0.0068 s_max 0.4955 s_min 0.2383
    [1][801] time: 37.27 a 0.6889 att_neg 3.0273 att_pos 0.9755 atte 1.7376 lr 0.0100 mls 0.5635 regu 0.1171 s_L2 0.0065 s_max 0.4773 s_min 0.2481
    [1][901] time: 37.34 a 0.7609 att_neg 2.9962 att_pos 1.0403 atte 1.8056 lr 0.0100 mls 0.6367 regu 0.1160 s_L2 0.0064 s_max 0.4861 s_min 0.2272
    Saving variables...
    --- cfp_fp ---
    testing verification..
    (14000, 96, 96, 3)
    # of images: 14000 Current image: 13952 Elapsed time: 00:00:12
    save /_feature.pkl
    sigma_sq (14000, 1)
    sigma_sq (14000, 1)
    sigma_sq [0.19821654 0.25770819 0.29024169 0.35030219 0.40342696 0.44539295
     0.56343746] percentile [0, 10, 30, 50, 70, 90, 100]
    risk_factor 0.0 risk_threshold 0.5634374618530273 keep_idxes 7000 / 7000 Cosine score acc 0.980429 threshold 0.182809
    risk_factor 0.1 risk_threshold 0.4627984762191772 keep_idxes 6301 / 7000 Cosine score acc 0.983336 threshold 0.201020
    risk_factor 0.2 risk_threshold 0.4453900158405304 keep_idxes 5603 / 7000 Cosine score acc 0.985007 threshold 0.203516
    risk_factor 0.3 risk_threshold 0.4327596127986908 keep_idxes 4904 / 7000 Cosine score acc 0.986134 threshold 0.207834
    

Testing

  • Single Image Comparison We use LFW dataset as an example for single image comparison. Make sure you have aligned LFW images using the previous commands. Then you can test it on the LFW dataset with the following command:
    run_eval.bat

Visualization of Uncertainty

Pre-trained Model

ResFace64

Method Download2 Download2
Base Mode Baidu Drive PW:v800 [Google Drive]TODO
MLS Only Baidu Drive PW:72tt [Google Drive]TODO
MLS + L1 + Triplet Baidu Drive PW:sx8a [Google Drive]TODO
ProbFace Baidu Drive PW:pr0m [Google Drive]TODO

ResFace64(0.5)

Method Download2 Download2
Base Mode Baidu Drive PW:zrkl [Google Drive]TODO
MLS Only Baidu Drive PW:et0e [Google Drive]TODO
MLS + L1 + Triplet Baidu Drive PW:glmf [Google Drive]TODO
ProbFace Baidu Drive PW:o4tn [Google Drive]TODO

Test Results:

Method LFW CFP-FF CALFW AgeDB30 CPLFW CFP-FP Vgg2FP Avg
Base Mode 99.80 99.80 95.93 97.93 92.53 98.04 94.92 96.99
MLS Only 99.80 99.76 95.87 97.35 93.01 98.29 95.26 97.05
MLS + L1 + Triplet 99.85 99.83 96.05 97.93 93.17 98.39 95.36 97.22
ProbFace 99.85 99.80 96.02 97.90 93.53 98.41 95.34 97.26

Acknowledgement

This repo is inspired by Probabilistic-Face-Embeddings

Reference

If you find this repo useful, please consider citing:

@misc{chen2021reliable,
    title={Reliable Probabilistic Face Embeddings in the Wild},
    author={Kai Chen and Qi Lv and Taihe Yi and Zhengming Yi},
    year={2021},
    eprint={2102.04075},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}
Owner
Kaen Chan
Kaen Chan
A python3 tool to take a 360 degree survey of the RF spectrum (hamlib + rotctld + RTL-SDR/HackRF)

RF Light House (rflh) A python script to use a rotor and a SDR device (RTL-SDR or HackRF One) to measure the RF level around and get a data set and be

Pavel Milanes (CO7WT) 11 Dec 13, 2022
Official implementation for "Low-light Image Enhancement via Breaking Down the Darkness"

Low-light Image Enhancement via Breaking Down the Darkness by Qiming Hu, Xiaojie Guo. 1. Dependencies Python3 PyTorch=1.0 OpenCV-Python, TensorboardX

Qiming Hu 30 Jan 01, 2023
CM-NAS: Cross-Modality Neural Architecture Search for Visible-Infrared Person Re-Identification (ICCV2021)

CM-NAS Official Pytorch code of paper CM-NAS: Cross-Modality Neural Architecture Search for Visible-Infrared Person Re-Identification in ICCV2021. Vis

JDAI-CV 40 Nov 25, 2022
Towards Improving Embedding Based Models of Social Network Alignment via Pseudo Anchors

PSML paper: Towards Improving Embedding Based Models of Social Network Alignment via Pseudo Anchors PSML_IONE,PSML_ABNE,PSML_DEEPLINK,PSML_SNNA: numpy

13 Nov 27, 2022
the code used for the preprint Embedding-based Instance Segmentation of Microscopy Images.

EmbedSeg Introduction This repository hosts the version of the code used for the preprint Embedding-based Instance Segmentation of Microscopy Images.

JugLab 88 Dec 25, 2022
Fast, differentiable sorting and ranking in PyTorch

Torchsort Fast, differentiable sorting and ranking in PyTorch. Pure PyTorch implementation of Fast Differentiable Sorting and Ranking (Blondel et al.)

Teddy Koker 655 Jan 04, 2023
For IBM Quantum Challenge 2021 (May 20 - 26)

IBM Quantum Challenge 2021 Introduction Commemorating the 40-year anniversary of the Physics of Computation conference, and 5-year anniversary of IBM

Qiskit Community 140 Jan 01, 2023
Tutorial on scikit-learn and IPython for parallel machine learning

Parallel Machine Learning with scikit-learn and IPython Video recording of this tutorial given at PyCon in 2013. The tutorial material has been rearra

Olivier Grisel 1.6k Dec 26, 2022
Code for the KDD 2021 paper 'Filtration Curves for Graph Representation'

Filtration Curves for Graph Representation This repository provides the code from the KDD'21 paper Filtration Curves for Graph Representation. Depende

Machine Learning and Computational Biology Lab 16 Oct 16, 2022
(JMLR'19) A Python Toolbox for Scalable Outlier Detection (Anomaly Detection)

Python Outlier Detection (PyOD) Deployment & Documentation & Stats Build Status & Coverage & Maintainability & License PyOD is a comprehensive and sca

Yue Zhao 6.6k Jan 03, 2023
Conceptual 12M is a dataset containing (image-URL, caption) pairs collected for vision-and-language pre-training.

Conceptual 12M We introduce the Conceptual 12M (CC12M), a dataset with ~12 million image-text pairs meant to be used for vision-and-language pre-train

Google Research Datasets 226 Dec 07, 2022
Keras implementations of Generative Adversarial Networks.

This repository has gone stale as I unfortunately do not have the time to maintain it anymore. If you would like to continue the development of it as

Erik Linder-Norén 8.9k Jan 04, 2023
Neural Articulated Radiance Field

Neural Articulated Radiance Field NARF Neural Articulated Radiance Field Atsuhiro Noguchi, Xiao Sun, Stephen Lin, Tatsuya Harada ICCV 2021 [Paper] [Co

Atsuhiro Noguchi 144 Jan 03, 2023
Code for paper: "Spinning Language Models for Propaganda-As-A-Service"

Spinning Language Models for Propaganda-As-A-Service This is the source code for the Arxiv version of the paper. You can use this Google Colab to expl

Eugene Bagdasaryan 16 Jan 03, 2023
Efficient 3D human pose estimation in video using 2D keypoint trajectories

3D human pose estimation in video with temporal convolutions and semi-supervised training This is the implementation of the approach described in the

Meta Research 3.1k Dec 29, 2022
This repository is to support contributions for tools for the Project CodeNet dataset hosted in DAX

The goal of Project CodeNet is to provide the AI-for-Code research community with a large scale, diverse, and high quality curated dataset to drive innovation in AI techniques.

International Business Machines 1.2k Jan 04, 2023
[CVPR2021] DoDNet: Learning to segment multi-organ and tumors from multiple partially labeled datasets

DoDNet This repo holds the pytorch implementation of DoDNet: DoDNet: Learning to segment multi-organ and tumors from multiple partially labeled datase

116 Dec 12, 2022
This repo provides the source code & data of our paper "GreaseLM: Graph REASoning Enhanced Language Models"

GreaseLM: Graph REASoning Enhanced Language Models This repo provides the source code & data of our paper "GreaseLM: Graph REASoning Enhanced Language

137 Jan 02, 2023
Source code for models described in the paper "AudioCLIP: Extending CLIP to Image, Text and Audio" (https://arxiv.org/abs/2106.13043)

AudioCLIP Extending CLIP to Image, Text and Audio This repository contains implementation of the models described in the paper arXiv:2106.13043. This

458 Jan 02, 2023
Training Certifiably Robust Neural Networks with Efficient Local Lipschitz Bounds (Local-Lip)

Training Certifiably Robust Neural Networks with Efficient Local Lipschitz Bounds (Local-Lip) Introduction TL;DR: We propose an efficient and trainabl

17 Dec 01, 2022