Auterion’s Post

New Embedded Vision Summit Demo Video: "INUITIVE Demonstration of the M4.51 Depth and AI Sensor Module Based on the NU4100 Vision Processor"

Autonomous vision-based low-altitude pipe tracking by multicopter: The validation of fully autonomous vision-based low-altitude pipe tracking has been done in GAZEBO simulation. The pipe geometry with various joints has been created in a simulation environment to test the performance of the algorithm during the turn.  Milestone completed: -> Development of vision strategy for vision-based control of multicopter. -> Implementation of homographic transformation to improve the tracking performance. ->Design and heuristically tuning of sliding mode controller. To test the algorithm in a real scenario the Convolution neural network has been designed by amalgamation of DeepLabv3+ and Squeeze Net. The model has been deployed on the NVIDIA Jetson Xavier-NX.

Digital twins are going to be essential for scaling simulation to scale AI for use cases that interact with humans and the real world environment. The automotive industry is pioneering the capture and use of data on and adjacent to public roads, and we'll soon see the same piece of road captured and simulated many times, and then briefly recaptured as vehicles pass through in the real world. I wonder if there's a use case for identifying the changes in digital twins based on local capture and measurement? This would enable the underlying environments used to simulate to be kept up to date, and to also underpin prediction models on environment changes. #digitaltwin #simulation #artificialintelligence

embedded world: Altera optimizes FPGAs for edge AI - Embedded.com https://buff.ly/43VrrQY

🚀 Exciting Breakthrough in LiDAR-Based Joint Detection and Tracking! 🚀 We're thrilled to share our latest research from the University of Toronto that has been accepted to #ECCV2024: "JDT3D: Addressing the Gaps in LiDAR-Based Tracking-by-Attention" by Brian Cheong, Jiachen (Jason) Zhou and Steven Lake Waslander. In computer vision, approaches trained in an end-to-end manner have been shown to perform better than traditional pipeline-based methods. However, within LiDAR-based object tracking, tracking-by-detection continues to achieve state-of-the-art performance without learning both tasks jointly. In our work, we explore the potential reasons for this gap and propose techniques to leverage the advantages of joint detection and tracking. 🌟 Key Highlights: - Innovative Approach: We propose JDT3D, a novel LiDAR-based joint detection and tracking model that leverages transformer-based decoders to propagate object queries over time, implicitly performing object tracking without an association step at inference. - Enhanced Techniques: We introduce track sampling augmentation and confidence-based query propagation to bridge the performance gap between tracking-by-detection (TBD) and joint detection and tracking (JDT) methods. - Real-World Impact: Our model is trained and evaluated on the nuScenes dataset, showcasing significant improvements in tracking accuracy and robustness. Check out the full paper linked below and join us at #ECCV2024! Paper: https://lnkd.in/gbT4EStA Code: https://lnkd.in/gU385pcW #autonomousdriving #computervision #tracking #objecttracking #robotics #3DVision #transformers #deeplearning

🎉 Excited to share that the work we have been working on for the past two years got accepted to #ECCV2024, the European Conference on Computer Vision. Our work #JDT3D (joint 3D object detection and tracking) aims to address the gaps in the current #Transformer based end-to-end detection and tracking architecture for #autonomousdriving. When I started my master, I strongly believed that object detection shouldn’t be done just on single frames, but should leverage the past frames of detection as much as possible while performing object tracking jointly in an implicit way. In this work, we took inspiration from the #Transformer (the now famous architecture in natural language processing), adopted it to 3D #LiDAR-based computer vision, and proposed several key optimization and training techniques (track sampling augmentation, confidence-based query propagation) to bridge the performance gap with the traditional two-step data association approaches. This work wouldn’t have been possible without the diligence and preseverance of Brian Cheong who is still working on further improvements in this area, and the unwavering support and supervision from Professor Steven Lake Waslander. As a commitment to the research community, we are preparing to open-source our work, so stay tuned by following us at Toronto Robotics and Artificial Intelligence Laboratory. Check out the full paper in detaills linked below. Paper arxiv link: https://lnkd.in/eVEBWU7S #autonomousdriving #computervision #lidar #objectdetection #multiobjecttracking #robotics #artificialintelligence

Please follow ProHawk AI on LinkedIn and help us get the word out on AI-enabled computer vision!

Wonders and Dangers of new tech https://lnkd.in/gSe7ejzH     With the type of scientific breakthroughs almost daily, we do not need sci-fi We are living in sci-fi become real world!   ‘Scientists have created jellyfish-shaped robots that can be steered through underwater obstacle courses using only light. This research has the potential to revolutionize #drug delivery systems and fluid manipulation in lab-on-a-chip devices.’ On the other hand, we also hear NHS UK is unable to see patients, as due to MS breakdown, patients records can’t be retrieved and people are stranded all over the world with flight cancellations.   As Technology takes over the world, are we becoming slaves to it?   Also if few lines of code from Crowdstrike (what a name!) software can do this, what happens to the world if #Internet is down for a few moments-leave alone days! With hackers adopting AI faster than us, is this doomsday near us?   Is that the biggest #Blackswan (Grey Rhino?) existential threat facing us all!  

AntiFake, Your Shield Against Synthesized Speech The AntiFake tool, created by the McKelvey School of Engineering at Washington University in St. Louis, serves as an innovative defense mechanism specifically engineered to prevent unauthorized speech synthesis proactively. Learn More about #AI Tech:  https://o-trim.co/ecohub X: https://o-trim.co/XTwitter https://lnkd.in/gRXRADiT #ONPASSIVE #ONPASSIVEECOSYSTEM #Techtrend #AI #research #robotics #technology #technews #TheFutureOfInternet #speechsynthesis

🚗 Exciting Breakthrough in Vehicle Side Slip Angle Estimation! 🚀 Do you ever wonder if there could be a side slip angle estimator for vehicles that is both accurate and robust for different vehicles? If yes, you'll be happy to know that our recent Open Access paper in the IEEE Transactions on Intelligent Transportation Systems answers this and more. We have made an exciting breakthrough in vehicle side slip angle estimation that you wouldn't want to miss! Check out our paper for more details. 🌐 Link to Paper: https://lnkd.in/dEFB-rRD Our work develops Kinematics-Structured Neural Networks for accurate and robust vehicle side slip angle estimation. What sets our approach apart? - Generalization capacity. - Sample-efficiency. - Physical explainability. - Seamless deployment with standard automotive hardware and sensors. We meticulously designed our networks using rigid-body kinematic principles. This resulted in a highly accurate estimator even with different vehicles and in varying conditions. Our approach is validated using an open race-car dataset, outperforming existing estimators with better accuracy and generalization capacity. #autonomousvehicles #autonomousdriving #neuralnetworks #sideslipestimation #stateestimation #lateralvelocityestimation #vehicledynamics