Tag Archives: emotions

VIA Smart Recognition Module Recognizes Emotion, Face, Age & Gender

VIA Technologies, a global leader in the development of embedded platform and system solutions have announced its VIA Smart Recognition Platform powered by the Qualcomm Snapdragon 820 embedded platform. The VIA Recognition platform offers a robust development suite for Vision-based applications with the possibility of developing applications that can be used for detecting facial expression, faces, age and even the gender type of an individual. The VIA Technologies announced the VIA Smart Recognition Platform during the Embedded World 2018.

The Platform provides support for object and facial recognition, emotion detection, age and gender detection. Apart from that, the platform can be used for people counting and tracking. Applications for VIA Recognition Platform can be found in office buildings for staff access control and be delivering personalized adverts in retail signage kiosks.  Others areas of note can be in traffic monitoring (counting of vehicles), security screening in airports and VIP places, payment authentication, surveillance and others. The platform’s advanced AI algorithm ensures speed and accuracy.

VIA_SOM-9X20_module Carrier Board

The VIA Smart Recognition platform is based on its VIA SOM-9×20 module powered by the Qualcomm Snapdragon 820 and coupled with a SOM-DB2 evaluation board. The significant difference with this current setup and the earlier SOM module is the pre-loaded facial and object recognition stack. The SOM-9×20 module features a Qualcomm Snapdragon 820 with 4x Cortex-A72-like cores Kryo cores, two of those run at 2.2GHz and the last two at 1.6GHz. Measuring at 82mm x 45mm, the module features a 64GB eMMc Flash memory, 4GB LPDDR4 SDRAM, a 624MHZ Adreno 530 GPU, and a Hexagon 680 DSP. It offers rich I/P and display interface through its MXM 3.0 314-pin connector including USB 3.0, USB 2.0, HDMI 2.0, SDIO, PCIe, MIPI CSI, MIPI DSI, and multi-function pins for UART, I2C, SPI, and GPIO. It comes with a wireless module that includes Wi-Fi 802.11 a/b/g/n/ac, Bluetooth 4.1 and a separate GPS/GNSS RF receiver with an antenna connector.

The carrier (SOM-DB2 expansion) I/O board is expected to help in accelerating system development. The board extends out the module interfaces, including 2x Ethernet 2x USB 3.0, micro-USB 2.0, HDMI, and a microSD. The VIA Smart Recognition Platform comes with a BSP that features Android 7.1.1 as well as the VIA Smart ETK (Embedded Tool Kit) comprising some APIs, including Watchdog Timer (WDT) for safeguarding against system crashes, GPIO access, RTC for auto-power on, and a sample app. A BSP supporting Linux Kernel 3.18.44 is also under development.

According to Richard Brown, VP International Marketing of VIA Technologies, “Computer vision technologies such as facial and object recognition are becoming a vital tool for boosting public safety and convenience everywhere from check-in counters and security screening in airports to self-service kiosks and payment authentication systems in supermarkets. The VIA Smart Recognition Platform speeds up the development and deployment of tailor-made systems that harness these cutting-edge technologies to deliver innovative new customer services and experiences.”

The VIA SOM-9X20 module and SOM-DB2 evaluation board are available now on VIA Technologies site, but you have to contact them for the prices. For inspiration on usage in computer vision, check out the VIA Smart Recognition Turnkey System brochure and VIA SOM-9X20 Machine Vision Platform brochure.

Deltu, An Interactive Delta Robot

Robots intelligence is going beyond borders and it may outsmart humans in some common games. Right now, these robots have their own personalities and if you are not skilled enough, they might get upset with you!

The interactive Delta three-arm robot, Deltu, is able to interact with humans. This Interactive design consists of 3 arms, an Arduino mind, a ‘personality’, and two iPads that run Unity3d applications. A HTTP request is send to the computer by the human’s application and then a Python server sends strings and commands to Arduino for controlling the robot.


Deltu uses three different applications using symmetry as an interpretation, a mirror and a reflection of our own image. The first game “Together” is a drawing game where Deltu imitates and interprets what we draw. The second one is a battle between the machine and the human. And the final third one is a memory game where the human must learn from the robot. These games were designed to emphasize the special relationship with robots and its evolution.


Deltu uses two iPads to play mimicking games with a human opponent. This sounds amazing – until you know that Deltu is very demanding. Once the opponent makes a mistake, it may stop and take selfies or browse apps. The robot creates patterns by pressing particular tiles on an iPad. If the movements are not mimed precisely, it will shake its adjoined arms in a side-to-side motion that appears to simulate frustration. Then, the robot exits the game and opens the camera app instead.It snaps a few photos of itself and uploads these to Instagram. It even takes a few moments to browse the Explore section and follow a new account, and looks through SoundCloud.

To understand how the robot would work, check it in action

According to the creators, the project explores the relationship between humans and artificial intelligence, as the role of these systems in society has not yet been defined.

This relationship is not only making our performance better, also it may become a source of learning. Currently, this project only supports iOS and works on iPads.

Deltu Interactive designer, Alexia Léchot, had also created other interactive games during her graduate study in ECAL – University of Art and Design in Lausanne, Switzerland.

Via : Arduino Blog