OmniML at Hannover Messe Showcasing the Power of Advanced Computer Vision Models on Autonomous Mobile Robots (AMR)
OmniML is joining Qualcomm at the Hannover Messe expo booth in Hall 5, Stand E34 in Germany for the week of May 30th to June 2nd. OmniML built a monocular depth estimation model that was ported onto an autonomous mobile robot (AMR) on top of Qualcomm’s Robotics RB5 Platform.
Depth Estimation Demo
The depth estimation model provides a spatial view of the AMR’s surroundings. It empowered the robot to better understand its 3D surroundings by providing a representation of the spatial structure of objects nearby. The model can recognize surrounding three-dimensional objects and gain recognition of the different field of view depths of the surrounding objects. For the demo, the AMR is stimulated to be in a warehouse setting with nearby boxes, cones, and other commonly found warehouse objects.
The image on the left represents a video camera recording live and on the right is the depth estimation spatial view as observed by the AMR robot using OmniML’s depth estimation model. OmniML and Qualcomm will be demoing the robot at Hannover Messe in a replicated warehouse setting while showcasing the robot’s point of view live on the big screen TV.
OmniML builds software to help customers design and train optimal models tailored to target hardware platforms that enable 20% to 5x faster models with higher accuracy and a 10x reduction in engineering and deployment efforts. OmniML was able to train, optimize, test, deploy, and operationalize the depth estimation model on the AMR unit in less than 2 weeks using our software toolset. OmniML’s software automates model design optimization with an elastic model infrastructure. It covers all the popular neural network architectures and is extensible to new model architectures.
Depth Estimation Usages
The potential use cases for this are near limitless. OmniML provides the capability to build, optimize, and deploy advanced CV models on any hardware device which will open the possibilities for robotic devices to become more capable and be more situationally aware. Good depth estimation is a crucial prerequisite to building 3D-aware AI/ML. No additional hardware is needed as the AI software is used to perceive depth using video images. This will enable multiple different types of self-contained robotic devices like AMRs, drones, AV delivery robots, fulfillment robots, etc. to perform more AI functionalities with less energy.
Beyond robotic use cases, good depth estimation plus hardware-aware optimization will enable better accuracy for autonomous driving cars, superior navigation (robots, cars, drones, etc.) in crowded space, a wider range of uses for smart cameras, and finer robotic arm maneuverability and object handling.
OmniML aims to effortlessly empower edge AI everywhere and help our customers operationalize hardware-aware models with increased speed, accuracy, and time to deployment. We make machine learning models smaller and faster which can drastically reduce neural network size without losing accuracy. For those that can be there in person, please come and check us out at Hannover Messe for the week of May 30th to June 2nd. We will also be showcasing our capabilities via a series of video clips and updates for the conference virtually.
Please follow us and stay up to date on new OmniML developments via our LinkedIn page: https://www.linkedin.com/company/omniml/ and check out our website to learn more https://omniml.ai/.
