Retail giants are investing heavily into replacing the kiosks and information booths with interactive kiosks. The kiosks have significantly modernized from a stand-alone robotics capability to offer a wide variety of functions including:
Virtual Assistants
A virtual shopping assistant can help the user with the purchase by providing personalized product recommendations. AI-powered chatbots can work as virtual assistants by providing additional product information, delivery timelines on out-of-stock products, color, and size recommendations.
Audience Profiling
Smart kiosks can have AI-enabled sensors and cameras that can help with face detection and profiling of kiosk users. It can also provide insights into user behavior and buying patterns.
Targeted Promotions
Advertisements, discount offers, and relevant information on the kiosk screen based on the viewer profile.
Advanced cloud computing and Artificial Intelligence (AI) capabilities have made these functionalities possible. To further reduce the response time and data bandwidth constraints, newer platform ad processors are considered that can provide AI inferencing on the local kiosk itself. Earlier, the sensor/camera data was captured and sent over to the Cloud to be analyzed by AI trained neural network. With advances in compute technology, “Inferencing at the Edge” is possible, where edge devices are capable to run an optimized AI model at the point of data capture.
Based on a similar concept, eInfochips developed a proof-of-concept of a smart remote kiosk-based on the Qualcomm QCS610 processor. The demo application was successfully showcased at the Amazon Web Services (AWS) Reinvent. The application can be used for different use cases, such as on-demand virtual assistance at a retail store or an automatic video call to a concierge for virtual support when a guest approaches a kiosk. It can also be used at a retail store where the kiosk can be used for virtual assistance, customer footfall analysis, and targeted promotions.
The main aim of these solutions is to showcase the compute capabilities of the Qualcomm QCS610 processor to enable artificial intelligence on a device at the network edge. It will reduce the need for network round trips for processing.
Key Features of Qualcomm QCS 610 processor
- Octa-core Kryo 460 CPU cores optimized for power.
- Dual 14-bit Qualcomm Spectra™ 250L Image Signal Processor capable of supporting dual sensors.
- Adreno 612 GPU with 64-bit addressing @ up to 845MHz with latest API support.
- 24 MP @ 30 fps with dual ISPs; each ISP is capable of 16 MP.
- Premium 4K @30fps H.264/ H.265 video capture and playback.
- Qualcomm AI Engine designed to support on-device machine learning.
eInfochips’ Camera RDK (Reference Design Kit) is based on a powerful octa-core QCS610 SoC. It is used at the kiosk as a camera-enabled smart console with intelligence at the edge for advanced AI and machine learning capabilities.
Key Camera RDK Features
- Dual ISP and 4K Ultra-HD, multiple camera support.
- Multiple wireless connectivity options.
- Display support; microphone and speaker support.
- Low latency, lean memory footprint.
- Pre-built software package support.
- Video analytics algorithms.
- In-built AI for face recognition, object detection, object tracking, and people counting.
- API support for functionalities like camera configuration, image quality, video capture, streaming, encoding, user, and event management.
We have ported a reusable, platform-agnostic camera framework (RCF) that is leveraged for video capture, face detection, and audio-video streaming. RCF can be used as a firmware platform for IP camera design and enables purpose-built core services to be reused to kick-start camera development. eInfochips has successfully deployed this framework for multiple customer projects.
Leveraging the strong experience on the Cloud connectivity, the solution also uses AWS IoT core to generate alerts, and Amazon Kinesis Video Streams (KVS) to start a video call with the concierge. The AWS IoT core enables secure connection for edge devices to the Cloud. It supports the latest M2M communication protocols, including Wi-Fi and LoRaWAN. It also ensures secure device connections with mutual authentication and end-to-end data encryption.
Amazon Kinesis Video Streams offers secure video streaming from connected devices to AWS cloud infrastructure. It encrypts, indexes, and stores videos that can be accessed through the APIs for playback, on-demand viewing, and post-processing for the web-based dashboard and vision-based decision making. The API also supports the open-source plug-in, WebRTC, for real-time media streaming across connected devices, web browsers, and mobile applications.
eInfochips has leveraged this camera framework for multiple customer projects. The Camera RDK reduces time and effort drastically by leveraging eInfochips’ Reusable Camera Framework (RCF) that can be used as a firmware platform for IP camera design. It is purpose-built to kick-start camera development, thus, allowing the OEMs to focus on developing and rolling out break-through technology innovations.
To know more about the solution, please reach out to marketing@einfochips.com.
