Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Apr 25, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Efficient event-based object detection with lightweight CNNs: a surprising advantage over transformers.

Sen Zhang, Pengfei Wang, Yaofeng Cheng

|April 24, 2026

Summary

This summary is machine-generated.

Related Concept Videos

Force Classification

Force Classification

Forces play a crucial role in the study of physics and engineering. They are essential in describing the motion, behavior, and equilibrium of objects in the physical world. Forces can be classified based on their origin, type, and direction of action.
Contact and non-contact forces are two of the most widely used categories of forces. As the name suggests, contact forces require physical contact between two objects to act upon each other. Examples of contact forces include frictional,...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Deep-Learning-Enhanced Bioimaging Via Energy Traps Regulated Lanthanide Nanoparticles.

Angewandte Chemie (International ed. in English)·2026

Same author

Nationwide occurrence and site-scale dynamics of antibiotics in landfill-associated waters in China.

Water research·2026

Same author

Effects of interlayer cations and hydration on PFAS adsorption and mobility in montmorillonite: A molecular dynamics study.

Journal of contaminant hydrology·2026

Same author

Soft tactile chip with in-situ sensing for haptic rendering and reverse feedback enhanced gross to fine teleoperation.

Nature communications·2026

Same author

Influence of foreleg microstructure and nanomechanical behavior on torsional performance in the transient landing impact of the Chinese stag beetle, Dorcus hopei Saunders.

Acta biomaterialia·2026

Same author

Human-Robot Interaction Strategy of Service Robot with Insufficient Capability in Self-Service Shop.

Biomimetics (Basel, Switzerland)·2026

Same journal

Multifunctional reconfigurable terahertz metasurface based on vanadium dioxide phase transition: achieving broadband absorption and efficient polarization conversion.

Applied optics·2026

Same journal

High-Q-factor electromagnetically induced transparency utilizing quasi-bound states in the continuum in an all-dielectric terahertz metasurface.

Applied optics·2026

Same journal

Automated stitching interferometry for high-precision metrology of X-ray mirrors.

Applied optics·2026

Same journal

Experimental demonstration of an approach to designing a metal-dielectric DBR resonant cavity structure.

Applied optics·2026

Same journal

High-precision wavefront reconstruction from a single-shot interferogram using a physics-driven hybrid feature calibration network.

Applied optics·2026

Same journal

Ultra-high-Q Fano resonance based on coupled topological corner states in Kagome photonic crystals.

Applied optics·2026

See all related articles

This study introduces EvConvLite, a lightweight CNN model for event camera object detection. EvConvLite achieves faster inference and higher accuracy than transformer models, improving robotic perception in dynamic environments.

Area of Science:

Robotics
Computer Vision
Artificial Intelligence

Background:

Robotic perception in dynamic environments is challenged by RGB camera limitations like latency and motion blur.
Event cameras offer high temporal resolution and dynamic range, but algorithms are underdeveloped.
Current object detection models struggle to balance accuracy and low latency for event cameras.

Purpose of the Study:

To develop a novel, lightweight Convolutional Neural Network (CNN)-based object detection model for event cameras.
To overcome the limitations of existing transformer and traditional CNN approaches in speed and accuracy.
To enhance robotic perception capabilities in highly dynamic environments.

Main Methods:

Proposed EvConvLite, a modernized, lightweight CNN architecture inspired by ConvNeXt.

Related Experiment Videos

Last Updated: Apr 25, 2026

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Author Spotlight: Enhancement of Salient Object Detection for Smart Grid Applications

Published on: December 15, 2023

Implemented strategic architectural improvements to traditional CNNs.

Evaluated the model on event camera datasets for object detection tasks.

Main Results:

EvConvLite achieved a 3x inference speedup compared to existing models.
The proposed model surpassed state-of-the-art transformer-based models in detection accuracy.
Demonstrated superior performance for event camera object detection.

Conclusions:

Modernized CNNs, like EvConvLite, can outperform transformers for event camera object detection.
EvConvLite offers a promising solution for accurate and low-latency robotic perception.
This work represents a significant advancement in event camera-based computer vision.