Ken Hayes’ Post

Join RB Sloan, David Swank, and team in San Antonio for #TechAdvantage24 in booth 1729 for an exclusive glimpse into the latest technologies, solutions, and strategies tailored for electric cooperative leaders. Discover how these innovations can revolutionize co-op efficiency, elevate performance, and refine strategic planning. Don't miss out on this opportunity to stay ahead of the curve ➡ See live demonstrations of interoperability in action. ➡ Experience the resiliency of DERs during power outage scenarios ➡ Participate in real-time simulations #utilities #resilience #sustainability #enxchange

Don't miss out on this exciting event! Register below to join the UAS team in Denver, May 1-2, to support our partner, Flyability, at the North American Launch Event of the new E3 UT Payload! This is an invaluable opportunity for our customers to deepen their knowledge, network with industry peers, and see the latest technological advancements in action. Why attend? Dive Deep into UT (Ultrasonic Thickness Measurement): Uncover how UT technology is revolutionizing inspections across various sectors. Expert Insights & User Experiences: Learn from Flyability experts and current UT users about the transformative impact of this technology. Practical Demonstrations: Get hands-on experience with the Elios 3 and its UT Payload, and understand why it's the perfect tool for your inspection needs. Spaces are limited – don't wait to secure your spot for an event that promises to enhance your understanding and capabilities in commercial inspection applications. Register Here: https://lnkd.in/eYRXz8EN? Reach out to us directly with any questions or for assistance with registration at UAS@gresco.com or (478) 315-0812. #FlyabilityElios3 #UTLaunchEvent

The Dragons Den in the MAGPIE project was a great method to have the demo and tool leads present themselves! 🔎 And to look beyond the demo implementation and tool development towards scaling! 📋 The Dragons from the involved ports - Port of Rotterdam, HAROPA PORT, DeltaPort GmbH & Co. KG, APS - Administração dos Portos de Sines e do Algarve, S.A. and EFIP (European Federation of Inland Ports) - did an amazing job 💡 Questions on the technologies, the business case, the link to different contexts and when implementation could already start Great to host the session & work towards a Handbook that capture the lessons learned on the demo's and tools

CyberDemo Augmenting Simulated Human Demonstration for Real-World Dexterous Manipulation We introduce CyberDemo, a novel approach to robotic imitation learning that leverages simulated human demonstrations for real-world tasks. By incorporating extensive data augmentation in a simulated environment, CyberDemo outperforms traditional in-domain real-world demonstrations when transferred to the real world, handling diverse physical and visual conditions. Regardless of its affordability and convenience in data collection, CyberDemo outperforms baseline methods in terms of success rates across various tasks and exhibits generalizability with previously unseen objects. For example, it can rotate novel tetra-valve and penta-valve, despite human demonstrations only involving tri-valves. Our research demonstrates the significant potential of simulated human demonstrations for real-world dexterous manipulation tasks. paper page: https://lnkd.in/gAsBJHQK

📃Scientific paper: CyberDemo: Augmenting Simulated Human Demonstration for Real-World Dexterous Manipulation Abstract: We introduce CyberDemo, a novel approach to robotic imitation learning that leverages simulated human demonstrations for real-world tasks. By incorporating extensive data augmentation in a simulated environment, CyberDemo outperforms traditional in-domain real-world demonstrations when transferred to the real world, handling diverse physical and visual conditions. Regardless of its affordability and convenience in data collection, CyberDemo outperforms baseline methods in terms of success rates across various tasks and exhibits generalizability with previously unseen objects. For example, it can rotate novel tetra-valve and penta-valve, despite human demonstrations only involving tri-valves. Our research demonstrates the significant potential of simulated human demonstrations for real-world dexterous manipulation tasks. More details can be found at https://lnkd.in/eZEtteTn Continued on ES/IODE ➡️ https://etcse.fr/KTxgZ ------- If you find this interesting, feel free to follow, comment and share. We need your help to enhance our visibility, so that our platform continues to serve you.

It's a next level of Art

Our paper accepted at #CVPR2024 is on arxiv now: https://lnkd.in/gpUU9x_q The paper is titled "Visual Prompting for Generalized Few-shot Segmentation: A Multi-scale Approach" and is co-authored with Mennatullah Siam and my supervisors Leonid Sigal and Jim Little. In this work, we proposed an approach to Generalized Few-shot Segmentation where we: - Prompt DETR style frameworks at multiple scales through learnable visual prompts, each representing a category, learned from few-shot demonstrations. - Refining the visual prompts through a causal novel-to-base prompt attention mechanism that disentangles the base visual prompts learned with abundant data from novel prompts to improve novel class accuracy. - Furthermore, we demonstrate an approach to fine-tune these learned visual prompts in a transductive manner, utilizing unlabelled query images in addition to few-shot support images to improve novel class performance and retain the base class performance.

We made a cool new demo during the hackathon at Founders, Inc. The robot does a real-world task from manufacturing - "testing incoming parts". In the demo, the robot picks up a bulb from a random position, then re-grabs it for accurate insertion into the tester, and then sorts working bulbs from non-working ones. Combining tasks like grabbing from a random position with accurate insertion is hard. So we used three small models, and this trick solved the problem. Meanwhile, training the robot is still quite simple. You have to record just 30 minutes of human demonstrations. (10 mins per model).