Leveraging Generative AI Synthetic and Social Media Data for Content Generalizability to Overcome Data Constraints in Vision Deep Learning DOI Creative Commons

Panteha Alipour,

Erika Gallegos

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 28, 2024

Abstract Generalizing deep learning models across diverse content types is a persistent challenge in domains like Facial Emotion Recognition (FER), where datasets often fail to reflect the wide range of emotional responses triggered by different stimuli. This study addresses issue generalizability comparing FER model performance between trained on video data collected controlled laboratory environment, extracted from social media platform (YouTube), and synthetic generated using Generative Adversarial Networks. The videos focus facial reactions advertisements, integration these sources seeks address underrepresented advertisement genres, reactions, individual diversity. Our leverage Convolutional Neural Networks Xception architecture, which fine-tuned category based sampling. ensures training validation represent categories, while testing includes novel evaluate rigorously. Precision-recall curves ROC-AUC metrics are used assess performance. Results indicate 7% improvement accuracy 12% increase precision-recall AUC when combining real-world data, demonstrating reduced overfitting enhanced generalizability. These findings highlight effectiveness integrating build systems that perform reliably more representative content.

Language: Английский

Leveraging generative AI synthetic and social media data for content generalizability to overcome data constraints in vision deep learning DOI Creative Commons
Panteha Alipour, Erika Gallegos

Artificial Intelligence Review, Journal Year: 2025, Volume and Issue: 58(5)

Published: Feb. 24, 2025

Language: Английский

Citations

0
Precision Psychiatry for Obsessive-Compulsive Disorder: Clinical Applications of Deep Learning Architectures DOI Open Access
Brian A. Zaboski, Lora Bednarek

Journal of Clinical Medicine, Journal Year: 2025, Volume and Issue: 14(7), P. 2442 - 2442

Published: April 3, 2025

Obsessive-compulsive disorder (OCD) is a complex psychiatric condition characterized by significant heterogeneity in symptomatology and treatment response. Advances neuroimaging, EEG, other multimodal datasets have created opportunities to identify biomarkers predict outcomes, yet traditional statistical methods often fall short analyzing such high-dimensional data. Deep learning (DL) offers powerful tools for addressing these challenges leveraging architectures capable of classification, prediction, data generation. This brief review provides an overview five key DL architectures-feedforward neural networks, convolutional recurrent generative adversarial transformers-and their applications OCD research clinical practice. We highlight how models been used the predictors response, diagnose classify OCD, advance precision psychiatry. conclude discussing implementation DL, summarizing its advances promises underscoring field.

Language: Английский

Citations

0
Leveraging Generative AI Synthetic and Social Media Data for Content Generalizability to Overcome Data Constraints in Vision Deep Learning DOI Creative Commons

Panteha Alipour,

Erika Gallegos

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 28, 2024

Abstract Generalizing deep learning models across diverse content types is a persistent challenge in domains like Facial Emotion Recognition (FER), where datasets often fail to reflect the wide range of emotional responses triggered by different stimuli. This study addresses issue generalizability comparing FER model performance between trained on video data collected controlled laboratory environment, extracted from social media platform (YouTube), and synthetic generated using Generative Adversarial Networks. The videos focus facial reactions advertisements, integration these sources seeks address underrepresented advertisement genres, reactions, individual diversity. Our leverage Convolutional Neural Networks Xception architecture, which fine-tuned category based sampling. ensures training validation represent categories, while testing includes novel evaluate rigorously. Precision-recall curves ROC-AUC metrics are used assess performance. Results indicate 7% improvement accuracy 12% increase precision-recall AUC when combining real-world data, demonstrating reduced overfitting enhanced generalizability. These findings highlight effectiveness integrating build systems that perform reliably more representative content.

Language: Английский

Citations

0