Bookmark and Share

2018 Vision Product of the Year Winners

The Vision Product of the Year Awards  are open to members of the Embedded Vision Alliance and recognize the innovation and achievement of the industry’s leading technology, service and end-product companies who are enabling the next generation of practical applications for computer vision.

2018 award winners were announced at the Santa Clara Convention Center during the Embedded Vision Summit! Click here to see what they and the media had to say!


8tree – dentCHECK®
8tree makes dentCHECK®, a 3D optical scanner (vision system) that is a purpose-built tool for the aviation maintenance industry. dentCHECK has been approved by Airbus, backed by Boeing & is being adopted by leading commercial airlines and maintenance, repair and overhaul (MRO) organizations. Additionally, aircraft OEMs are now implementing dentCHECK for airframe inspections on the flightline. The tool is delivering >90% efficiency gains vs. traditional manual measurement methods.

Winner in the category of End Products


AImotive – aiSIM
In self-driving car development, testing hazardous or dangerous traffic situations is almost impossible and extremely costly without simulation technology. For AI-based technologies, this virtual environment provides a platform for training and testing, accelerating their development. aiSim offers a complete, integrated simulation environment, fine-tuned for a vision-first approach to autonomous driving through photorealistic rendering, sensor modeling with importable calibration parameters, dataflow simulation and software-in-the-loop/hardware-in-the-loop testing. This allows for the reliable assessment of image processing algorithms. These capabilities are extended with the integration of various virtual sensors, standardized map data, importable scenarios and procedural evaluation.

Winner in the categories of Processors and Developer Tools

AImotive – aiWare
aiWare is a highly-optimized hardware architecture designed for generic artificial neural network acceleration. The design is optimized for high-performance in power-constrained environments, such as self-driving, mobile devices or VR headsets. Scalable from embedded solutions to data centers, aiWare is engineered to facilitate the real-world use of artificial intelligence. Incorporating our team's experience with high-resolution input recognition tasks in self-driving, aiWare is designed to handle complex neural networks working on continuous high-resolution data-streams at 30 frames per second. The architecture provides high performance in latency critical environments, with unparalleled power efficiency.

Winner in the categories of Best Processor


Algolux – Algolux CANA
CANA is a novel end-to-end deep neural network (DNN) for much more robust perception in challenging imaging conditions, such as low light, adverse weather, or lens issues found in the real world. RAW or lightly processed image sensor data is directly input into the CANA DNN, bypassing traditional ISP pre-processing. Lens and sensor characteristics are fused into the early levels of the DNN to understand how the image is being formed, while the upper layers address the perception task, such as object detection, classification, or segmentation. The network is trained end-to-end with a tailored training approach that leverages a combination of captured and synthetic annotated data. This results in significantly improved perception in difficult imaging conditions, typically 20-30% versus current state of the art and higher in more extreme scenarios.

Winner in the category of Automotive Solutions


Intel – Intel RealSense
Intel RealSense Depth Cameras: D415 and D435 Intel® RealSense- D400 family of Depth cameras and modules brings enhanced 3D perception to more devices and machines that only see 2D today. The D400 family offers turnkey solutions for rapid product development and integration for VR, Robotics, and other markets where depth in computer vision matters. The backbone of this solution is our RealSense Vision Processor that uses advanced stereo algorithms to compute high resolution, high frame rate real time depth at throughputs of up to 36.6MP/s, without burdening the compute resources of a host GPU or CPU. RealSense Depth Modules are a collection of camera sensors packed in compact, pre-calibrated form factors for easy system integration. RealSense Depth Cameras combines our depth modules, Processor intelligence and software into one neatly packaged product ready out of the box for developers, makers, and innovators to get started with application development.

Winner in the category of Cameras, Modules, Sensors


Mathworks – GPU Coder
The new MathWorks® GPU Coder software enables scientists and engineers to automatically generate optimized CUDA code from high-level functional descriptions in MATLAB® for deep learning, embedded vision, and autonomous systems. The generated CUDA code, integrated in projects as source code or libraries, accelerates computationally intensive portions of the MATLAB code for modern GPUs including the NVIDIA Tesla®, embedded NVIDIA Tegra® System-on-Chip (SoC), NVIDIA Jetson- System-on-Modules (SoMs), and NVIDIA DRIVE- platforms. This automated workflow provides easy access to GPUs without requiring expert knowledge of GPU programming.

Winner in the category of Software and Algorithms


Morpho – SoftNeuro
SoftNeuro is one of the world's fastest deep learning inference engines. It operates in multiple environments, utilizing learning results that have been obtained through a variety of deep learning frameworks. SoftNeuro is not limited to image recognition, but can be used as a general-purpose inference engine for text analysis, voice recognition, image recognition, etc. It obtains profile data from the target platforms that execute inference, and performs optimizations based in the data to achieve higher speeds, and support inference processing on trained networks from all the mainstream frameworks, making it easy to deploy trained networks.

Winner in the category of AI Technology


Xilinx – Xilinx Machine Learning Suite
The Xilinx Machine Learning Suite provides tools for accelerating vision applications in the cloud. The key innovation of the Xilinx Machine Learning Suite is that it enables cloud users of machine learning inference to get an order of magnitude performance advantage/cost savings of optimized FPGA acceleration over GPUs; without the effort of developing a custom FPGA accelerator. Other innovations are the unique fixed-point quantization, layer fusing and FPGA memory optimizations that help maximize the acceleration benefits of the FPGA compute performance. The suite delivers lower-latency, higher-throughput, lower-cost machine learning inference in real world

Winner in the category of Cloud Technologies