Advances in Event Vision and Object Detection

The field of event vision and object detection is rapidly evolving, with a focus on improving the efficiency and accuracy of event filtering and object detection algorithms. Recent developments have led to the creation of innovative hardware architectures and software frameworks that enable high-throughput event processing and robust object detection in various environments. Notably, researchers are exploring the use of wavelet denoising, bidirectional guided low-light image enhancement, and hierarchical neural collapse detection to address challenges such as noise accumulation, low-light conditions, and catastrophic forgetting. These advancements have significant implications for applications in robot perception, surveillance, and autonomous systems. Noteworthy papers include: High Throughput Event Filtering, which proposes a hardware architecture for the Distance-based Interpolation with Frequency Weights filter, achieving a throughput of 403.39 million events per second. Bidirectional Image-Event Guided Low-Light Image Enhancement, which introduces a framework for low-light image enhancement using event guidance, outperforming state-of-the-art methods by 0.96dB in PSNR. WD-DETR, which proposes a wavelet denoising-enhanced real-time object detection transformer for event cameras, achieving a high frame rate of approximately 35 FPS.

Sources

High Throughput Event Filtering: The Interpolation-based DIF Algorithm Hardware Architecture

Bidirectional Image-Event Guided Low-Light Image Enhancement

Hierarchical Neural Collapse Detection Transformer for Class Incremental Object Detection

Data Augmentation For Small Object using Fast AutoAugment

WD-DETR: Wavelet Denoising-Enhanced Real-Time Object Detection Transformer for Robot Perception with Event Cameras

CEM-FBGTinyDet: Context-Enhanced Foreground Balance with Gradient Tuning for tiny Objects

Built with on top of