Workshop on Fair, Data efficient, and Trusted Computer Vision

FOR MISSING KEYNOTE VIDEOS BELOW, PLEASE SEE HERE INSTEAD (ALL VIDEOS AVAILABLE).
PLEASE SEE HERE FOR FULL PROGRAM. THIS IS AN ALTERNATE SITE FOR ALL PROGRAM CONTENT (INCLUDING SLIDES AND VIDEOS).

Teaser picture for paper
Alex Pentland
    Authors: Alex Pentland   
    Keywords:  fairness, legality
Sund Jun14  
9:00 AM - 9:25 AM
Favorite
Teaser picture for paper
An Analytical Framework for Trusted Machine Learning and Computer Vision Running with Blockchain Tao Wang, Maggie Du, Xinmin Wu, Taiping He SA
    Authors: Tao X Wang; Maggie Du; Xinmin Wu; Taiping He    
    Keywords:  machine learning, blockchain, computer vision, trust, astore
Sund Jun14  
9:25 AM - 9:30 AM
Favorite
Teaser picture for paper
In many areas in machine learning, decision trees play a crucial role in classification and regression. When a decision tree based classifier is hoste
    Authors: Kanthi K Sarpatwar ; Nalini Ratha ; Karthik Nandakumar ; Karthikeyan Shanmugam ; James Rayfield ; sharath pankanti ; Roman Vaculin    
    Keywords:  Decision Tree Classification Privacy Preserving Inference Homomorphic Encryption Packing Mechanisms Approximate Comparisons
Sund Jun14  
9:30 AM - 9:40 AM
Favorite
Teaser picture for paper
We derive a protocol to perform frequency-domain homomorphic convolution bewtween an encrypted vector and a plaintext vector and achieve 11x speedup.
    Authors: Song Bian Tianchen Wang, Masayuki Hiromoto Yiyu Shi, Takashi Sato   
    Keywords:  homomorphic encryption, secure inference, discrete Fourier transform, homomorphic convolution, neural networks, secret sharing, number theoretic transform, privacy, security
Sund Jun14  
9:40 AM - 9:50 AM
Favorite
Teaser picture for paper
Deep encoder/decoder lets you use regular RGB images like QR codes: hide a message, print/display the encoded img, take a photo, and recover the data!
    Authors: Matt Tancik,Ben Mildenhall,Ren Ng    
    Keywords:  steganography, camera, printed, hidden, qr
Sund Jun14  
9:50 AM - 10:00 AM
Favorite
Teaser picture for paper
Decision-making in Robotics with Vision-in-the-Loop: Best Practices and Open Problems
    Authors: Debadeepta Dey   
    Keywords:  robotics, UAVs, drones, anytime, neural, network, pipeline, optimization, airsim, simulation
Sund Jun14  
10:00 AM - 10:30 AM
Favorite
Teaser picture for paper
Manolis Savva
    Authors: Manolis Savva   
    Keywords:  trust, data efficient
Sund Jun14  
10:35 AM - 11:05 AM
Favorite
Teaser picture for paper
We propose an unsupervised method for learning weakly symmetric deformable 3D objects from raw single-view images, without any additional supervision.
    Authors: Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi   
    Keywords:  3D Reconstruction, Deformable 3D Objects, Unsupervised Learning, Intrinsic Image Decomposition, Single Image 3D, Unsupervised 3D Reconstruction
Sund Jun14  
11:05 AM - 11:15 AM
Favorite
Teaser picture for paper
LeGR makes automated filter pruning efficient to obtain multiple pruned networks on the accuracy/FLOPs trade-off curve with SoTA pruning accuracy.
    Authors: Ting-Wu(Rudy)Chin, RuizhouDing, Cha Zhang, Diana Marculescu   
    Keywords:  Model Compression, Filter Pruning, AutoML, Image Classification
Sund Jun14  
11:15 AM - 11:25 AM
Favorite
Teaser picture for paper
Our aim is to minimize the supervision required for multi-label categorization. Specifically, we investigate an effective class of approaches that ass
    Authors: Rajat ; Munender Varshney ; Pravendra Singh ; Vinay P Namboodiri    
    Keywords:  Active Learning, Multi-Label Categorization, Weakly Supervised Learning, Localization based catogrization, Deep Learning.
Sund Jun14  
11:25 AM - 11:30 AM
Favorite
Teaser picture for paper
Fisher Yu
    Authors: Fisher Yu   
    Keywords:  trust, data efficient
Sund Jun14  
11:30 AM - 12:00 PM
Favorite
Teaser picture for paper
Tal Hassner
    Authors: Tal Hassner   
    Keywords:  bias, fairness
Sund Jun14  
1:00 PM - 1:30 PM
Favorite
Teaser picture for paper
We propose a style-based face aging framework to transfer multi-scale aging patterns. The method is used to enhance the diversity of facial datasets.
    Authors: Markos Georgopoulos ; James A Oldfield Mihalis A Nicolaou; Yannis Panagakis ; Maja Pantic    
    Keywords:  age progression, dataset bias, diversity, style transfer
Sund Jun14  
1:30 PM - 1:40 PM
Favorite
Teaser picture for paper
Adversarial representation learning is a promising paradigm for obtaining data representations that are invariant to certain sensitive attributes whil
    Authors: Bashir Sadeghi ; Vishnu Boddeti    
    Keywords:  Fairness, Pre-Trained, Kernelization, Closed-Form Solution, Optimal Dimensionality
Sund Jun14  
1:40 PM - 1:50 PM
Favorite
Teaser picture for paper
A novel adversarial derived data augmentation methodology to mitigate racial bias of face recognition based on the cyclic adversarial training.
    Authors: Seyma Yucer ; Samet Akcay ; Noura Al Moubayed; Toby Breckon    
    Keywords:  face recognition, visual recognition bias, generative adversarial networks, data augmentation, computer vision
Sund Jun14  
1:50 PM - 2:00 PM
Favorite
Teaser picture for paper
Christian Theobalt
    Authors: Christian Theobalt   
    Keywords:  trust, data efficient
Sund Jun14  
2:00 PM - 2:30 PM
Favorite
Teaser picture for paper
Hany Farid
    Authors: Hany Farid   
    Keywords:  trust, data efficient
Sund Jun14  
2:30 PM - 3:00 PM
Favorite
Teaser picture for paper
We present a novel, parameter-free “defense layer” which blocks the generation of adversarial noise without adding any computational overhead.
    Authors: Akhil Goel ; Akshay Agarwal ;Mayank Vatsa ; Richa Singh ; Nalini Ratha    
    Keywords:  Adversarial Machine Learning, Adversarial Robustness, CNN, Backpropagation, Image Classification
Sund Jun14  
3:00 PM - 3:05 PM
Favorite
Teaser picture for paper
Want to train robust models with less computational overhead? Use Plug And Pipeline (PAP), an efficient regularizer for adversarial training.
    Authors: Vivek B S ; Ambareesh Revanur ; Naveen Venkat ; Venkatesh Babu RADHAKRISHNAN    
    Keywords:  Adversarial training, Robustness, Efficient training, Computational efficiency, Representation learning
Sund Jun14  
3:05 PM - 3:15 PM
Favorite
Teaser picture for paper
Adversarial training shows large generalization error. In our work, we propose Adversarial Vertex mixup as the solution to the generalization problem
    Authors: Saehyung Lee Hyungyu Lee Sungroh Yoon   
    Keywords:  adversarial training, adversarially robust generalization, mixup, adversarial defense, adversarial examples
Sund Jun14  
3:15 PM - 3:25 PM
Favorite
Teaser picture for paper
We propose U-Net GAN, which uses a U-Net shaped discriminator for detailed per-pixel feedback and improves over the SOTA BigGAN baseline.
    Authors: Edgar Schönfeld, Bernt Schiele, Anna Khoreva   
    Keywords:  GAN, image synthesis, U-Net, discriminator, consistency regularization
Sund Jun14  
3:25 PM - 3:35 PM
Favorite
Teaser picture for paper
    Authors: All authors   
Sund Jun14  
3:40 PM - 4:40 PM
Favorite
Teaser picture for paper
Alex Pentland
    Authors: Alex Pentland   
    Keywords:  fairness, legality
Sund Jun14  
9:00 PM - 9:25 PM
Favorite
Teaser picture for paper
An Analytical Framework for Trusted Machine Learning and Computer Vision Running with Blockchain Tao Wang, Maggie Du, Xinmin Wu, Taiping He SA
    Authors: Tao X Wang; Maggie Du; Xinmin Wu; Taiping He    
    Keywords:  machine learning, blockchain, computer vision, trust, astore
Sund Jun14  
9:25 PM - 9:30 PM
Favorite
Teaser picture for paper
In many areas in machine learning, decision trees play a crucial role in classification and regression. When a decision tree based classifier is hoste
    Authors: Kanthi K Sarpatwar ; Nalini Ratha ; Karthik Nandakumar ; Karthikeyan Shanmugam ; James Rayfield ; sharath pankanti ; Roman Vaculin    
    Keywords:  Decision Tree Classification Privacy Preserving Inference Homomorphic Encryption Packing Mechanisms Approximate Comparisons
Sund Jun14  
9:30 PM - 9:40 PM
Favorite
Teaser picture for paper
We derive a protocol to perform frequency-domain homomorphic convolution bewtween an encrypted vector and a plaintext vector and achieve 11x speedup.
    Authors: Song Bian Tianchen Wang, Masayuki Hiromoto Yiyu Shi, Takashi Sato   
    Keywords:  homomorphic encryption, secure inference, discrete Fourier transform, homomorphic convolution, neural networks, secret sharing, number theoretic transform, privacy, security
Sund Jun14  
9:40 PM - 9:50 PM
Favorite
Teaser picture for paper
Deep encoder/decoder lets you use regular RGB images like QR codes: hide a message, print/display the encoded img, take a photo, and recover the data!
    Authors: Matt Tancik,Ben Mildenhall,Ren Ng    
    Keywords:  steganography, camera, printed, hidden, qr
Sund Jun14  
9:50 PM - 10:00 PM
Favorite
Teaser picture for paper
Decision-making in Robotics with Vision-in-the-Loop: Best Practices and Open Problems
    Authors: Debadeepta Dey   
    Keywords:  robotics, UAVs, drones, anytime, neural, network, pipeline, optimization, airsim, simulation
Sund Jun14  
10:00 PM - 10:30 PM
Favorite
Teaser picture for paper
Manolis Savva
    Authors: Manolis Savva   
    Keywords:  trust, data efficient
Sund Jun14  
10:35 PM - 11:05 PM
Favorite
Teaser picture for paper
We propose an unsupervised method for learning weakly symmetric deformable 3D objects from raw single-view images, without any additional supervision.
    Authors: Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi   
    Keywords:  3D Reconstruction, Deformable 3D Objects, Unsupervised Learning, Intrinsic Image Decomposition, Single Image 3D, Unsupervised 3D Reconstruction
Sund Jun14  
11:05 PM - 11:15 PM
Favorite
Teaser picture for paper
LeGR makes automated filter pruning efficient to obtain multiple pruned networks on the accuracy/FLOPs trade-off curve with SoTA pruning accuracy.
    Authors: Ting-Wu(Rudy)Chin, RuizhouDing, Cha Zhang, Diana Marculescu   
    Keywords:  Model Compression, Filter Pruning, AutoML, Image Classification
Sund Jun14  
11:15 PM - 11:25 PM
Favorite
Teaser picture for paper
Our aim is to minimize the supervision required for multi-label categorization. Specifically, we investigate an effective class of approaches that ass
    Authors: Rajat ; Munender Varshney ; Pravendra Singh ; Vinay P Namboodiri    
    Keywords:  Active Learning, Multi-Label Categorization, Weakly Supervised Learning, Localization based catogrization, Deep Learning.
Sund Jun14  
11:25 PM - 11:30 PM
Favorite
Teaser picture for paper
Fisher Yu
    Authors: Fisher Yu   
    Keywords:  trust, data efficient
Sund Jun14  
June 14 - June 15
Favorite
Teaser picture for paper
Tal Hassner
    Authors: Tal Hassner   
    Keywords:  bias, fairness
Mond Jun15  
1:00 AM - 1:30 AM
Favorite
Teaser picture for paper
We propose a style-based face aging framework to transfer multi-scale aging patterns. The method is used to enhance the diversity of facial datasets.
    Authors: Markos Georgopoulos ; James A Oldfield Mihalis A Nicolaou; Yannis Panagakis ; Maja Pantic    
    Keywords:  age progression, dataset bias, diversity, style transfer
Mond Jun15  
1:30 AM - 1:40 AM
Favorite
Teaser picture for paper
Adversarial representation learning is a promising paradigm for obtaining data representations that are invariant to certain sensitive attributes whil
    Authors: Bashir Sadeghi ; Vishnu Boddeti    
    Keywords:  Fairness, Pre-Trained, Kernelization, Closed-Form Solution, Optimal Dimensionality
Mond Jun15  
1:40 AM - 1:50 AM
Favorite
Teaser picture for paper
A novel adversarial derived data augmentation methodology to mitigate racial bias of face recognition based on the cyclic adversarial training.
    Authors: Seyma Yucer ; Samet Akcay ; Noura Al Moubayed; Toby Breckon    
    Keywords:  face recognition, visual recognition bias, generative adversarial networks, data augmentation, computer vision
Mond Jun15  
1:50 AM - 2:00 AM
Favorite
Teaser picture for paper
Christian Theobalt
    Authors: Christian Theobalt   
    Keywords:  trust, data efficient
Mond Jun15  
2:00 AM - 2:30 AM
Favorite
Teaser picture for paper
Hany Farid
    Authors: Hany Farid   
    Keywords:  trust, data efficient
Mond Jun15  
2:30 AM - 3:00 AM
Favorite
Teaser picture for paper
We present a novel, parameter-free “defense layer” which blocks the generation of adversarial noise without adding any computational overhead.
    Authors: Akhil Goel ; Akshay Agarwal ;Mayank Vatsa ; Richa Singh ; Nalini Ratha    
    Keywords:  Adversarial Machine Learning, Adversarial Robustness, CNN, Backpropagation, Image Classification
Mond Jun15  
3:00 AM - 3:05 AM
Favorite
Teaser picture for paper
Want to train robust models with less computational overhead? Use Plug And Pipeline (PAP), an efficient regularizer for adversarial training.
    Authors: Vivek B S ; Ambareesh Revanur ; Naveen Venkat ; Venkatesh Babu RADHAKRISHNAN    
    Keywords:  Adversarial training, Robustness, Efficient training, Computational efficiency, Representation learning
Mond Jun15  
3:05 AM - 3:15 AM
Favorite
Teaser picture for paper
Adversarial training shows large generalization error. In our work, we propose Adversarial Vertex mixup as the solution to the generalization problem
    Authors: Saehyung Lee Hyungyu Lee Sungroh Yoon   
    Keywords:  adversarial training, adversarially robust generalization, mixup, adversarial defense, adversarial examples
Mond Jun15  
3:15 AM - 3:25 AM
Favorite
Teaser picture for paper
We propose U-Net GAN, which uses a U-Net shaped discriminator for detailed per-pixel feedback and improves over the SOTA BigGAN baseline.
    Authors: Edgar Schönfeld, Bernt Schiele, Anna Khoreva   
    Keywords:  GAN, image synthesis, U-Net, discriminator, consistency regularization
Mond Jun15  
3:25 AM - 3:35 AM
Favorite
Teaser picture for paper
    Authors: All authors   
Mond Jun15  
3:40 AM - 4:40 AM
Favorite
Teaser picture for paper
Aleksandra Mojsilovic
    Authors: Aleksandra (Saška) Mojsilovic   
    Keywords:  fairness, trust, data efficient
Mond Jun15  
9:00 AM - 9:30 AM
Favorite
Teaser picture for paper
Attribution maps by three representative methods to explain the same prediction (match stick: 0.535) made by a ResNet-50 classifier to an ImageNet ima
    Authors: Naman Bansal ; Chirag Agarwal ; Anh Nguyen    
    Keywords:  Hyper-parameter Sensitivity, Robust Classifiers, Explanation maps or Attribution Maps or Heatmaps, Interpretability, Localization Error
Mond Jun15  
9:30 AM - 9:40 AM
Favorite
Teaser picture for paper
In this paper, we discuss the potential issues of the commonly used quality metrics for uncertainty estimation and propose new metrics to mitigate the
    Authors: Yukun Ding; Jinglan Liu ; Jinjun Xiong; Yiyu Shi    
    Keywords:  Uncertainty Estimation, Selective Prediction, Confidence Calibration, Trustworthiness
Mond Jun15  
9:40 AM - 9:50 AM
Favorite
Teaser picture for paper
Evaluating attribution methods with necessity and sufficiency, concepts which can be quantified with ordering and its refinement, proportionality.
    Authors: Zifan Wang ; Piotr Mardziel ; Matt Fredrikson ; Anupam Datta    
    Keywords:  Evaluation Criteria, Attribution Map, Interpretation, Necessity, Sufficiency
Mond Jun15  
9:50 AM - 10:00 AM
Favorite
Teaser picture for paper
Pushmeet Kohli
    Authors: Pushmeet Kohli   
    Keywords:  trust, data efficient
Mond Jun15  
10:00 AM - 10:30 AM
Favorite
Teaser picture for paper
Timnit Gebru
    Authors: Timnit Gebru   
    Keywords:  trust, data efficient
Mond Jun15  
10:30 AM - 11:00 AM
Favorite
Teaser picture for paper
In this work we investigate the reasons why images may fail to be classified correctly. We do this by understanding the importance of the filters in t
    Authors: Thomas Hartley ; Kirill Sidorov ; Chris Willis ; David Marshall    
    Keywords:  Explainable, XAI, Interpretable, Failure, Surprise
Mond Jun15  
11:00 AM - 11:10 AM
Favorite
Teaser picture for paper
In this paper, we develop a novel gradient-free visual explanation method named Score-CAM, which bridges the gap between perturbation-based and CAM-ba
    Authors: Haofan Wang ; Zifan Wang; Mengnan Du ; Fan Yang ; Zijian Zhang ; Sirui Ding ; Piotr Mardziel ; Xia Hu    
    Keywords:  Explainability, Visual explanation, XAI, Class activation map, Computer vision
Mond Jun15  
11:10 AM - 11:20 AM
Favorite
Teaser picture for paper
This work discusses the potential leak of one's biometric information via images posted on social media. While posting photos of themselves or highlig
    Authors: Aakarsh Malhotra ; Saheb Chhabra ; Mayank Vatsa ; Richa Singh    
    Keywords:  Identity anonymization, Finger-selfie, fingerprints, recognition, adversarial learning
Mond Jun15  
11:20 AM - 11:25 AM
Favorite
Teaser picture for paper
Tsung-Yi Lin
    Authors: Tsung-Yi Lin   
    Keywords:  trust, data efficient
Mond Jun15  
11:25 AM - 11:55 AM
Favorite
Teaser picture for paper
Hima Lakkaraju
    Authors: Hima Lakkaraju   
    Keywords:  trust, explanations
Mond Jun15  
1:00 PM - 1:30 PM
Favorite
Teaser picture for paper
Activity classification has observed great success recently. The performance on small dataset is almost saturated and people are moving towards larg
    Authors: Jialing Lyu ; Weichao Qiu; Alan Yuille    
    Keywords:  Activity Classification, Model Understanding, Synthetic Data
Mond Jun15  
1:30 PM - 1:40 PM
Favorite
Teaser picture for paper
We propose active learning method via self-supervised Fisher kernel to minimize labeling efforts by 40% for biased datasets with 10x less processing.
    Authors: Denis Gudovskiy Alec Hodgkinson Takuya Yamaguchi Sotaro Tsukizawa   
    Keywords:  active learning, data bias, class imbalance, self-supervised learning, Fisher kernel
Mond Jun15  
1:40 PM - 1:50 PM
Favorite
Teaser picture for paper
We correct an existing dataset and re-evaluate a model on the corrected corpus, SNLI-VE-2.0. Then we introduce e-SNLI-VE-2.0, which appends natural la
    Authors: Virginie Do ; Oana-Maria Camburu ; Zeynep Akata ; Thomas Lukasiewicz    
    Keywords:  multimodal classification, textual entailment, explanations, visual reasoning, natural language generation
Mond Jun15  
1:50 PM - 2:00 PM
Favorite
Teaser picture for paper
Matthias Niessner
    Authors: Matthias Nießner   
    Keywords:  trust, data efficient
Mond Jun15  
2:10 PM - 2:40 PM
Favorite
Teaser picture for paper
We present a new public experimental framework to study biases in multimodal machine learning. FairCVtest is inspired in automated recruitment tools.
    Authors: Alejandro Peña ; Ignacio Serna ;Aythami Morales ; Julian Fierrez   
    Keywords:  Multimodal AI, Automated Recruitment, Fairness, Algorithmic Discrimination, Computer Vision
Mond Jun15  
2:40 PM - 2:50 PM
Favorite
Teaser picture for paper
Recent incidents have highlighted the biased behavior (for or against a sub-group) of deep learning based automated models for different classificatio
    Authors: Shruti Nagpal ; Maneet Singh ; Richa Singh ; Mayank Vatsa    
    Keywords:  Bias-Invariant Classification, Unbiased Predictions, Face Attribute, Imbalanced Training, Fair AI
Mond Jun15  
2:50 PM - 3:00 PM
Favorite
Teaser picture for paper
Bias in FR is, rightfully so, at the attention of many, and researchers and consumers alike. We built a balanced face verification dataset as a proxy
    Authors: Joseph P Robinson ; YUN FU ; Yann Henon ; Gennady Livitz ; Can Qin ; Samson Timoner    
    Keywords:  Fairness, Facial Verification, Transparency in ML, Labeled Face Dataset, Trusting AI, Balanced Data
Mond Jun15  
3:00 PM - 3:05 PM
Favorite
Teaser picture for paper
    Authors: All authors   
Mond Jun15  
3:30 PM - 4:30 PM
Favorite
Teaser picture for paper
Aleksandra Mojsilovic
    Authors: Aleksandra (Saška) Mojsilovic   
    Keywords:  fairness, trust, data efficient
Mond Jun15  
9:00 PM - 9:30 PM
Favorite
Teaser picture for paper
Attribution maps by three representative methods to explain the same prediction (match stick: 0.535) made by a ResNet-50 classifier to an ImageNet ima
    Authors: Naman Bansal ; Chirag Agarwal ; Anh Nguyen    
    Keywords:  Hyper-parameter Sensitivity, Robust Classifiers, Explanation maps or Attribution Maps or Heatmaps, Interpretability, Localization Error
Mond Jun15  
9:30 PM - 9:40 PM
Favorite
Teaser picture for paper
In this paper, we discuss the potential issues of the commonly used quality metrics for uncertainty estimation and propose new metrics to mitigate the
    Authors: Yukun Ding; Jinglan Liu ; Jinjun Xiong; Yiyu Shi    
    Keywords:  Uncertainty Estimation, Selective Prediction, Confidence Calibration, Trustworthiness
Mond Jun15  
9:40 PM - 9:50 PM
Favorite
Teaser picture for paper
Evaluating attribution methods with necessity and sufficiency, concepts which can be quantified with ordering and its refinement, proportionality.
    Authors: Zifan Wang ; Piotr Mardziel ; Matt Fredrikson ; Anupam Datta    
    Keywords:  Evaluation Criteria, Attribution Map, Interpretation, Necessity, Sufficiency
Mond Jun15  
9:50 PM - 10:00 PM
Favorite
Teaser picture for paper
Pushmeet Kohli
    Authors: Pushmeet Kohli   
    Keywords:  trust, data efficient
Mond Jun15  
10:00 PM - 10:30 PM
Favorite
Teaser picture for paper
Timnit Gebru
    Authors: Timnit Gebru   
    Keywords:  trust, data efficient
Mond Jun15  
10:30 PM - 11:00 PM
Favorite
Teaser picture for paper
In this work we investigate the reasons why images may fail to be classified correctly. We do this by understanding the importance of the filters in t
    Authors: Thomas Hartley ; Kirill Sidorov ; Chris Willis ; David Marshall    
    Keywords:  Explainable, XAI, Interpretable, Failure, Surprise
Mond Jun15  
11:00 PM - 11:10 PM
Favorite
Teaser picture for paper
In this paper, we develop a novel gradient-free visual explanation method named Score-CAM, which bridges the gap between perturbation-based and CAM-ba
    Authors: Haofan Wang ; Zifan Wang; Mengnan Du ; Fan Yang ; Zijian Zhang ; Sirui Ding ; Piotr Mardziel ; Xia Hu    
    Keywords:  Explainability, Visual explanation, XAI, Class activation map, Computer vision
Mond Jun15  
11:10 PM - 11:20 PM
Favorite
Teaser picture for paper
This work discusses the potential leak of one's biometric information via images posted on social media. While posting photos of themselves or highlig
    Authors: Aakarsh Malhotra ; Saheb Chhabra ; Mayank Vatsa ; Richa Singh    
    Keywords:  Identity anonymization, Finger-selfie, fingerprints, recognition, adversarial learning
Mond Jun15  
11:20 PM - 11:25 PM
Favorite
Teaser picture for paper
Tsung-Yi Lin
    Authors: Tsung-Yi Lin   
    Keywords:  trust, data efficient
Mond Jun15  
11:25 PM - 11:55 PM
Favorite
Teaser picture for paper
Hima Lakkaraju
    Authors: Hima Lakkaraju   
    Keywords:  trust, explanations
Tues Jun16  
1:00 AM - 1:30 AM
Favorite
Teaser picture for paper
Activity classification has observed great success recently. The performance on small dataset is almost saturated and people are moving towards larg
    Authors: Jialing Lyu ; Weichao Qiu; Alan Yuille    
    Keywords:  Activity Classification, Model Understanding, Synthetic Data
Tues Jun16  
1:30 AM - 1:40 AM
Favorite
Teaser picture for paper
We propose active learning method via self-supervised Fisher kernel to minimize labeling efforts by 40% for biased datasets with 10x less processing.
    Authors: Denis Gudovskiy Alec Hodgkinson Takuya Yamaguchi Sotaro Tsukizawa   
    Keywords:  active learning, data bias, class imbalance, self-supervised learning, Fisher kernel
Tues Jun16  
1:40 AM - 1:50 AM
Favorite
Teaser picture for paper
We correct an existing dataset and re-evaluate a model on the corrected corpus, SNLI-VE-2.0. Then we introduce e-SNLI-VE-2.0, which appends natural la
    Authors: Virginie Do ; Oana-Maria Camburu ; Zeynep Akata ; Thomas Lukasiewicz    
    Keywords:  multimodal classification, textual entailment, explanations, visual reasoning, natural language generation
Tues Jun16  
1:50 AM - 2:00 AM
Favorite
Teaser picture for paper
Matthias Niessner
    Authors: Matthias Nießner   
    Keywords:  trust, data efficient
Tues Jun16  
2:10 AM - 2:40 AM
Favorite
Teaser picture for paper
We present a new public experimental framework to study biases in multimodal machine learning. FairCVtest is inspired in automated recruitment tools.
    Authors: Alejandro Peña ; Ignacio Serna ;Aythami Morales ; Julian Fierrez   
    Keywords:  Multimodal AI, Automated Recruitment, Fairness, Algorithmic Discrimination, Computer Vision
Tues Jun16  
2:40 AM - 2:50 AM
Favorite
Teaser picture for paper
Recent incidents have highlighted the biased behavior (for or against a sub-group) of deep learning based automated models for different classificatio
    Authors: Shruti Nagpal ; Maneet Singh ; Richa Singh ; Mayank Vatsa    
    Keywords:  Bias-Invariant Classification, Unbiased Predictions, Face Attribute, Imbalanced Training, Fair AI
Tues Jun16  
2:50 AM - 3:00 AM
Favorite
Teaser picture for paper
Bias in FR is, rightfully so, at the attention of many, and researchers and consumers alike. We built a balanced face verification dataset as a proxy
    Authors: Joseph P Robinson ; YUN FU ; Yann Henon ; Gennady Livitz ; Can Qin ; Samson Timoner    
    Keywords:  Fairness, Facial Verification, Transparency in ML, Labeled Face Dataset, Trusting AI, Balanced Data
Tues Jun16  
3:00 AM - 3:05 AM
Favorite
Teaser picture for paper
    Authors: All authors   
Tues Jun16  
3:30 AM - 4:30 AM
Favorite