Home › Categories › Technology› Download

Download.pdf

Factorized_Higher-Order_CNNs_for_Spatio-Temporal_Emotion_Estimation.pdf

🔗 Source: adrianbulat.com

📊 Size: 802 KB

👤 Author: Jean Kossaifi, Antoine Toisoul, Adrian Bulat, Yannis Panagakis, Timothy M. Hospedales, Maja Pan

⬇️ Downloads: 72

Summary

Researchers propose a tensor factorization framework for efficient multidimensional convolutions of higher-order, unifying approaches to compress networks and reduce parameters. The framework, called CP-Higher-Order Convolution (HO-CPConv), enables a novel higher-order transduction, allowing a network trained on 2D images to generalize to higher-order data like videos. Applied to spatio-temporal facial emotion analysis, the methodology demonstrates superior performance on three large-scale affect estimation datasets: AffectNet, SEWA, and AFEW-VA. The approach reduces parameters and improves computational efficiency, making it suitable for resource-limited platforms.

Download File

Description

Researchers propose a tensor factorization framework for efficient multidimensional convolutions of higher-order, unifying approaches to compress networks and...

Technical Information

File Format: PDF
File Size: 802 KB
Pages: 10
Language: EN
Author: Jean Kossaifi, Antoine Toisoul, Adrian Bulat, Yannis Panagakis, Timothy M. Hospedales, Maja Pan
Total Downloads: 72
Last Updated: 1 month ago

Document Overview

This PDF document about Download provides comprehensive information and guidance. Whether you're a beginner or advanced user, this resource offers valuable insights into Download.