Filters








13,953 Hits in 4.8 sec

Attribution-Driven Explanation of the Deep Neural Network Model via Conditional Microstructure Image Synthesis

Shusen Liu, Bhavya Kailkhura, Jize Zhang, Anna M. Hiszpanski, Emily Robertson, Donald Loveland, Xiaoting Zhong, T. Yong-Jin Han
2022 ACS Omega  
In this work, we propose a technique for interpreting the behavior of deep learning models by injecting domain-specific attributes as tunable "knobs" in the material optimization analysis pipeline.  ...  However, despite their effectiveness in building highly predictive models, e.g., predicting material properties from microstructure imaging, due to their opaque nature fundamental challenges exist in extracting  ...  ■ ACKNOWLEDGMENTS This work was performed under the auspices of the U.S.  ... 
doi:10.1021/acsomega.1c04796 pmid:35097261 pmcid:PMC8793074 fatcat:45tkiipic5fgdc544gqqv7ze34

Towards Explainable, Privacy-Preserved Human-Motion Affect Recognition [article]

Matthew Malek-Podjaski, Fani Deligianni
2021 arXiv   pre-print
In particular, we propose a cross-subject transfer learning technique for training a multi-encoder autoencoder deep neural network to learn disentangled latent representations of human motion features.  ...  However, gait is composed of many highly expressive characteristics that can be used to identify human subjects, and most solutions fail to address this, disregarding the subject's privacy.  ...  From Local to Global Explanations We are interested in evaluating interpretability across class samples in order to understand global properties of the prediction model.  ... 
arXiv:2105.03958v2 fatcat:er4dcdi3hrfzdb66qwz73yjhbu

Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations [article]

Jiaheng Wei, Zhaowei Zhu, Hao Cheng, Tongliang Liu, Gang Niu, Yang Liu
2022 arXiv   pre-print
study the property and treatment of real-world label noise; (2) These efforts are often of large scales, which may result in unfair comparisons of robust methods within reasonable and accessible computation  ...  To better understand real-world label noise, it is crucial to build controllable and moderate-sized real-world noisy datasets with both ground-truth and noisy labels.  ...  Our work contains the human subject study which involves only simply image annotation tasks in Amazon Mechanical Turk.  ... 
arXiv:2110.12088v2 fatcat:q57onvwrcbggzk5jwbzum7m7je

Stopping Criterion during Rendering of Computer-Generated Images Based on SVD-Entropy

Jérôme Buisine, André Bigand, Rémi Synave, Samuel Delepoulle, Christophe Renaud
2021 Entropy  
Until now, the features taking part in the human evaluation of image quality and the remaining perceived noise are not precisely known.  ...  This noise can be reduced by increasing the number of paths, as proved by Monte Carlo theory, but the problem of finding the right number of paths that are required in order to ensure that human observers  ...  From those subjective human thresholds, it is finally possible to provide a label (noisy or noiseless) to each sub-image of any 800 × 800 image: sub-images in a block whose sampling level is less than  ... 
doi:10.3390/e23010075 pmid:33419115 fatcat:6smznuevx5fbhikrsf62stxdvu

Dive into Ambiguity: Latent Distribution Mining and Pairwise Uncertainty Estimation for Facial Expression Recognition [article]

Jiahui She, Yibo Hu, Hailin Shi, Jun Wang, Qiu Shen, Tao Mei
2021 arXiv   pre-print
Due to the subjective annotation and the inherent interclass similarity of facial expressions, one of key challenges in Facial Expression Recognition (FER) is the annotation ambiguity.  ...  For the former, an auxiliary multi-branch learning framework is introduced to better mine and describe the latent distribution in the label space.  ...  Acknowledgements This work is supported by the National Key R&D Program of China under Grant No. 2020AAA0103800 and by the National Natural Science Foundation of China under Grant U1936202 and 62071216  ... 
arXiv:2104.00232v1 fatcat:nhrxl6ix5jes5ekcp5bt23twp4

A Survey on Bias in Visual Datasets [article]

Simone Fabbrizzi, Symeon Papadopoulos, Eirini Ntoutsi, Ioannis Kompatsiaris
2021 arXiv   pre-print
Thus, both the problems of understanding and discovering biases are of utmost importance. Yet, to date there is no comprehensive survey on bias in visual datasets.  ...  Computer Vision (CV) has achieved remarkable results, outperforming humans in several tasks. Nonetheless, it may result in major discrimination if not dealt with proper care.  ...  In order to avoid the complexity of free text description, in the first step, workers are presented with a batch of images and asked to describe the similarities among the images of the batch via a question-answer  ... 
arXiv:2107.07919v1 fatcat:ibhsebbwwzaw3n5rdedmv3a6we

Understanding Workers, Developing Effective Tasks, and Enhancing Marketplace Dynamics: A Study of a Large Crowdsourcing Marketplace [article]

Ayush Jain, Akash Das Sarma, Aditya Parameswaran, Jennifer Widom
2017 arXiv   pre-print
Using this data---never before analyzed in an academic context---we shed light on three crucial aspects of crowdsourcing: (1) Task design --- helping requesters understand what constitutes an effective  ...  load; and (3) Worker behavior --- understanding worker attention spans, lifetimes, and general behavior, for the improvement of the crowdsourcing ecosystem as a whole.  ...  Almost 51,000, or 88% of the 58,000 of batches have some representatives in our 12,000 batch sample-thus, the sample is missing about 10% of the tasks.  ... 
arXiv:1701.06207v1 fatcat:pzhkid4nivevtbk4b4hlq2nmei

Curriculum Learning of Visual Attribute Clusters for Multi-Task Classification [article]

Nikolaos Sarafianos, Theodore Giannakopoulos, Christophoros Nikou, Ioannis A. Kakadiaris
2018 arXiv   pre-print
In this paper, we introduce a novel method to combine the advantages of both multi-task and curriculum learning in a visual attribute classification framework.  ...  Visual attributes, from simple objects (e.g., backpacks, hats) to soft-biometrics (e.g., gender, height, clothing) have proven to be a powerful representational approach for many applications such as image  ...  Acknowledgments This work has been funded in part by the UH Hugh Roy and Lillie Cranz Cullen Endowment Fund. The work of C.  ... 
arXiv:1709.06664v3 fatcat:b3i7y7x63nb7bjbaxedviwwmry

Image Aesthetic Assessment: An experimental survey

Yubin Deng, Chen Change Loy, Xiaoou Tang
2017 IEEE Signal Processing Magazine  
This survey aims at reviewing recent computer vision techniques used in the assessment of image aesthetic quality.  ...  Image aesthetic assessment aims at computationally distinguishing high-quality photos from low-quality ones based on photographic rules, typically in the form of binary classification or quality scoring  ...  Subjective evaluation by conducting human survey is also seen in [38] where human evaluators are asked to give subjective aesthetic attribute ratings in order to calibrate a proposed aesthetic signature  ... 
doi:10.1109/msp.2017.2696576 fatcat:eexxuo6d3rbndfvubemsd7mxtu

What do You Mean? Interpreting Image Classification with Crowdsourced Concept Extraction and Analysis

Agathe Balayn, Panagiotis Soilis, Christoph Lofi, Jie Yang, Alessandro Bozzon
2021 Proceedings of the Web Conference 2021  
In this paper, we introduce a scalable human-inthe-loop approach for global interpretability.  ...  Salient image areas identified by local interpretability methods are annotated with semantic concepts, which are then aggregated into a tabular representation of images to facilitate automatic statistical  ...  A closer look at Figure 4b shows that, once more, most incorrect concepts have Cramer's values inferior to 0.2 when increasing the number of images since such concepts are more subject to sampling noise  ... 
doi:10.1145/3442381.3450069 fatcat:pbmvaeysnrh3tiz3ziar26zfdm

Medical image editing in the latent space of Generative Adversarial Networks

Rubén Fernández, Pilar Rosado, Esteban Vegas, Ferran Reverter
2021 Intelligence-Based Medicine  
We analyze thousands of image patches from whole-slide images of breast cancer metastases in histological lymph node sections.  ...  We consider a set of arithmetic operations in the latent space of Generative Adversarial Networks (GANs) to edit histopathological images.  ...  This work was supported in part by a grant from the 2017 SGR 622 (Generalitat de Catalunya) and in part from the PID2019-104830RB-I00 (Ministerio de Ciencia e Innovación y Ministerio de Universidades),  ... 
doi:10.1016/j.ibmed.2021.100040 fatcat:hlnxg2nogzh7dburs3zeixadxm

Distraction-Aware Feature Learning for Human Attribute Recognition via Coarse-to-Fine Attention Mechanism [article]

Mingda Wu, Di Huang, Yuanfang Guo, Yunhong Wang
2019 arXiv   pre-print
Recently, Human Attribute Recognition (HAR) has become a hot topic due to its scientific challenges and application potentials, where localizing attributes is a crucial stage but not well handled.  ...  Extensive experiments are conducted on the WIDER-Attribute and RAP databases, and state-of-the-art results are achieved, demonstrating the effectiveness of the proposed approach.  ...  Input images (1st row) in HAR usually contain target subjects (masked in red) and distractions (enclosed in green contours).  ... 
arXiv:1911.11351v1 fatcat:wspdlr2vivf3zepgjp77dj4fim

Distraction-Aware Feature Learning for Human Attribute Recognition via Coarse-to-Fine Attention Mechanism

Mingda Wu, Di Huang, Yuanfang Guo, Yunhong Wang
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
Recently, Human Attribute Recognition (HAR) has become a hot topic due to its scientific challenges and application potentials, where localizing attributes is a crucial stage but not well handled.  ...  Extensive experiments are conducted on the WIDER-Attribute and RAP databases, and state-of-the-art results are achieved, demonstrating the effectiveness of the proposed approach.  ...  Input images (1st row) in HAR usually contain target subjects (masked in red) and distractions (enclosed in green contours).  ... 
doi:10.1609/aaai.v34i07.6925 fatcat:bvfau3wpibbafgvn37r7l3jo3y

Coming to Grips with Age Prediction on Imbalanced Multimodal Community Question Answering Data

Alejandro Figueroa, Billy Peralta, Orietta Nicolis
2021 Information  
Answers, user demographics have impacts on their revenues and user experience; demographics assist in ensuring that the needs of each cohort are fulfilled via personalizing and contextualizing content.  ...  As for textual inputs, we propose an age-batched greedy curriculum learning (AGCL) approach to lessen the effects of their inherent class imbalances.  ...  The label Virtual human indicates a virtual human avatar; otherwise it does not correspond to a virtual human avatar. Figure 13 . 13 Figure 13.  ... 
doi:10.3390/info12020048 fatcat:xvmng6a4yncgdfo5ef3xooeh3m

Hierarchical Variational Autoencoders For Visual Counterfactuals

Nicolas Vercheval, Aleksandra Pizurica
2021 2021 IEEE International Conference on Image Processing (ICIP)  
In this paper we show how relaxing the effect of the posterior leads to successful counterfactuals and we introduce VAEX 1 an Hierarchical VAE designed for this approach that can visually audit a classifier  ...  in applications.  ...  Unfortunately, this approach alone is not often successful, and has little impact on the reconstructed image.  ... 
doi:10.1109/icip42928.2021.9506780 fatcat:lybgy3e37zhvxei5eytfyi7gou
« Previous Showing results 1 — 15 out of 13,953 results