Lattice Semiconductor will deliver the free webinar “Architecting Always-On, Context-Aware, On-Device AI Using Flexible Low-power FPGAs” on October 30, 2018 at 9 am Pacific Time, in partnership with the Embedded Vision Alliance. The webinar will be presented by Deepak Boppana, the company’s Senior Director of Marketing, and Gordon Hands, Marketing Director for IP and Solutions (and a highly-rated Embedded Vision Summit presenter). In this webinar, the presenters will leverage the company’s experience in developing low-cost, low-power, always-on, vision-based AI solutions to illustrate deep learning inferencing design tradeoffs and explore optimizations across edge processing implementations ranging from 1 mW to 1 W and $1 to $10. For more information, and to register, see the event page.
The next session of the Embedded Vision Alliance’s in-person, hands-on technical training class series, Deep Learning for Computer Vision with TensorFlow, takes place in two weeks in San Jose, California. These classes give you the critical knowledge you need to develop deep learning computer vision applications with TensorFlow. The one-day class takes place on October 4, 2018. Details, including online registration, can be found here.
Brian Dipert Editor-In-Chief, Embedded Vision Alliance
HARDWARE AND SOFTWARE DEVELOPMENT FOR RESOURCE-CONSTRAINED SYSTEMS
Computer Vision for Augmented Reality in Embedded Designs Augmented reality (AR) and related technologies are becoming increasingly popular and prevalent, led by their adoption in smartphones, tablets and other mobile computing and communications devices. While developers of more deeply embedded platforms are also motivated to incorporate AR capabilities in their products, the comparative scarcity of processing, memory, storage, and networking resources is challenging, as are cost, form factor, power consumption and other constraints. However, by making effective use of available compute capabilities, along with APIs, middleware and other software toolsets, implementing robust AR in resource-constrained designs is increasingly feasible.
Building Efficient CNN Models for Mobile and Embedded Applications Recent advances in efficient deep learning models have led to many potential applications in mobile and embedded devices. In this talk, Peter Vajda, Research Scientist at Facebook, discusses state-of-the-art model architectures, and introduces Facebook’s work on real-time style transfer and pose estimation on mobile phones.
ADVANCED IMAGE SENSING AND PROCESSING TECHNIQUES
Generative Sensing: Reliable Recognition from Unreliable Sensor Data While deep neural networks (DNNs) perform on par with – or better than – humans on pristine high-resolution images, DNN performance is significantly worse than human performance on images with quality degradations, which are frequently encountered in real-world applications. This talk from Lina Karam, Professor and Computer Engineering Director at Arizona State University, introduces a new generative sensing framework which integrates low-end sensors with computational intelligence to attain recognition accuracy on par with that attained using high-end sensors. This generative sensing framework aims to transform low-quality sensor data into higher quality data in terms of classification accuracy. In contrast with existing methods for image generation, this framework is based on discriminative models and aims to maximize recognition accuracy rather than a similarity measure. This is achieved through the introduction of selective feature regeneration in a deep neural network.
Neuromorphic Event-based Computer Vision: Sensors, Theory and Applications In this presentation, Ryad B. Benosman, Professor at the University of Pittsburgh Medical Center, Carnegie Mellon University and Sorbonne Universitas, introduces neuromorphic, event-based approaches for image sensing and processing. State-of-the-art image sensors suffer from severe limitations imposed by their very principle of operation. These sensors acquire visual information as a series of “snapshots” recorded at discrete points in time, hence time-quantized at a predetermined frame rate, resulting in limited temporal resolution, low dynamic range and a high degree of redundancy in the acquired data. Nature suggests a different approach: Biological vision systems are driven and controlled by events happening within the scene in view, and not – like conventional image sensors – by artificially created timing and control signals that have no relation to the source of the visual information. Translating the frameless paradigm of biological vision to artificial imaging systems implies that control over the acquisition of visual information is no longer imposed externally on an array of pixels but rather the decision making is transferred to each individual pixel, which handles its own information individually. Benosman introduces the fundamentals underlying such bio-inspired, event-based image sensing and processing approaches, and explores their strengths and weaknesses. He shows that bio-inspired vision systems have the potential to outperform conventional, frame-based vision acquisition and processing systems and to establish new benchmarks in terms of data compression, dynamic range, temporal resolution and power efficiency in applications such as 3D vision, object tracking, motor control and visual feedback loops, in real-time.