Advances in Visual Recognition and Enhancement

The field of visual recognition and enhancement is moving towards more efficient and accurate models, with a focus on real-time performance and limited data requirements. Researchers are exploring new approaches to sign language recognition, salient object detection, and image enhancement, leveraging techniques such as transformer-based architectures, convolutional neural networks, and attention mechanisms. Notable papers in this area include: SignBart, which presents a novel approach to sign language recognition using an encoder-decoder architecture, achieving 96.04% accuracy on the LSA-64 dataset. Oneta, which proposes a multi-style image enhancement model using eigentransformation functions, demonstrating high performance on six enhancement tasks across 30 datasets. Rapid Salient Object Detection with Difference Convolutional Neural Networks, which introduces an efficient network design for salient object detection, achieving significant improvements in efficiency-accuracy trade-offs. MobileIE, which presents an extremely lightweight ConvNet for real-time image enhancement on mobile devices, achieving real-time inference at up to 1,100 frames per second.

Sources

SignBart -- New approach with the skeleton sequence for Isolated Sign language Recognition

A Novel Frame Identification and Synchronization Technique for Smartphone Visible Light Communication Systems Based on Convolutional Neural Networks

From Sight to Insight: Unleashing Eye-Tracking in Weakly Supervised Video Salient Object Detection

Oneta: Multi-Style Image Enhancement Using Eigentransformation Functions

Developing Lightweight DNN Models With Limited Data For Real-Time Sign Language Recognition

Rapid Salient Object Detection with Difference Convolutional Neural Networks

Exploring Pose-based Sign Language Translation: Ablation Studies and Attention Insights

MobileIE: An Extremely Lightweight and Effective ConvNet for Real-Time Image Enhancement on Mobile Devices

Flow-CDNet: A Novel Network for Detecting Both Slow and Fast Changes in Bitemporal Images

Built with on top of