BrainFrame Forum

BrainFrame, a powerful, open, smart vision platform

As a developer or system integrator of Intelligent Video Analytics (IVA) applications, are you concerned that you cannot find an easy-to-use and powerful IVA solution? Are you struggling with the uncertainty of the return on investment (ROI) of your smart vision IOT projects? As a developer of computer vision or deep learning algorithms, do you face the challenge where specific algorithms can be implemented easily but building a product around them is difficult? As a hardware manufacturer of smart cameras, are you worried about slow product deployment because there is no end-to-end software solution to integrate your product with?

A powerful, open, smart vision platform, BrainFrame, has officially launched! In early January 2020, Aotu announced BrainFrame and the OpenVisionCapsules standard, co-released with OpenCV.org, upon which BrainFrame is built. Users and developers can now download a free version of BrainFrame from Aotu’s website at aotu.ai/docs/downloads/.

Videos in, structured data out. Edge computing, plug-and-play, ready to use for various use cases.

  • BrainFrame runs video processing and AI inference on edge computing devices, extracting real-time insights from stationary cameras for various, continuous monitoring, tracking and analysis tasks, such as:
    • Recognizing and tracking people, counting, zoning, traffic, vehicles;
    • Classifying objects; behavior analysis;
    • Incident alerts, etc.
  • Structured data is output in the format of graphical reports, statistics reports and/or real-time alarms/action cards, allowing developers and system integrators to process multiple tasks for broad inspection & monitoring use cases.
  • BrainFrame comes with many out-of-the-box smart vision applications*, requiring no development, including:
    • Administration: Queue/occupancy, Shelf space/inventory management
    • Industrial: Production line defect management, Safety compliance management
    • Retail: Retail/restaurant, bank retail store management
    • Traffic/urban management & analysis, etc.
  • Users configure the regions, lines, alarms/action cards with UI and video overlay based on the business needs of the operation. Graphical report format or alarms/action cards can be adjusted on-the-go to fit users’ needs: aotu.ai/docs/dashboard/getting_started/

Fast to deploy, easy to use, ready to scale.

  • User-oriented, What It Sees Is What You Get (WISIWYG) smart vision platform. The user simply installs BrainFrame and it’s ready for use—no programming required. Following the instructions at aotu.ai/docs/user_guide/server_setup/, software installation on x86 computers takes about one hour. The system comes with a client, which supports Linux or Windows, a database dashboard, real-time alarms/action cards, and many more features. Simply enter the network addresses of the IP cameras, configure the areas of interest and alarms/action cards, and then start to use the system. The system configuration for cameras usually takes just a couple minutes aotu.ai/docs/user_guide/client_setup/.
  • If the computer has GPUs installed, any high performance, high accuracy VisionCapsules can be used; if there is no GPU, OpenVino compatible VisionCapsules can be used.
  • BrainFrame can be used as a standalone AI application for a few camera streams, or as an enterprise or industrial video analytics service with clustering and multi-premises* support for thousands of camera streams or more.

Easy to customize and integrate. BrainFrame comes with many vision applications—a user can install and immediately begin to use it with no programming required. It also has powerful support for customization to extend its capabilities to fit users’ unique needs.

  • If needed, Application Developers and System Integrators can develop BrainFrame applications based on REST API and database API: aotu.ai/docs/api/ to integrate BrainFrame in Enterprise or Industrial applications.
  • BrainFrame’s VisionCapsules utilize the OpenVisionCapsules standard. It allows a user to drag and drop algorithms in OpenVisionCapsules format to enable applications in Administration, Industrial, Retail, Traffic, etc.; algorithm developers can easily integrate their own VisionCapsules to build their Video AI applications. It usually takes less than 50 total lines of code to develop a new end-to-end VisionCapsule when referencing the sample code. OpenVisionCapsules is an open-source, BSD licensed system for encapsulation of machine learning and computer vision algorithms, which Aotu co-released with OpenCV. The OpenVisionCapsules source code is available on OpenCV’s GitHub at github.com/opencv/open_vision_capsules.

Patent-pending AI technologies and algorithms: BrainFrame automated algorithm fusion, integration, and inference optimization scheduling engine

  • BrainFrame’s AI inference pipeline and scheduler have been fully optimized for video streaming on x86 platforms and Nvidia GPU architectures.
  • BrainFrame is compatible with Intel CPUs, Intel Graphics, Intel Movidius/FPGA* and Nvidia GPU AI acceleration platforms. BrainFrame can support other silicon vendors’ solutions through its VisionCapsule approach.
  • Users can load or unload their choice of plug-and-play VisionCapsules depending on their use case. VisionCapsule loading and data/algorithm fusion is fully automated and happens at run-time. Loading/unloading has no impact on the business’ operation.
  • If a new vision algorithm or deep neural network needs to be developed for a unique use case, an algorithm developer can quickly repurpose published VisionCapsules, and focus their effort on development of the new algorithm. BrainFrame will automatically integrate the new algorithm with the existing algorithms, dynamically at run-time, connecting VisionCapsule together and forming a data-flow graph.
  • Fully functional UI to control VisionCapsule deployment—select which streams a capsule is activated for and control runtime parameters.

For Smart Camera and Smart Vision Hardware OEMs, and System Integrators:

  • Hardware manufactures can independently design BrainFrame-compatible smart cameras using the OpenVisionCapsules standard.
  • OEMs can preinstall BrainFrame software on smart vision edge computing devices to support enterprise applications or urban management applications.
  • System integrators can use BrainFrame + Non-smart-vision camera + smart vision cameras to deploy edge/device/cloud hybrid computing architecture, to build a solution maximizing computing power while minimizing bandwidth usage.

(* Note: Some functionality will be available to the public soon. Contact sales for early access)

BrainFrame Downloads and Documentation:

aotu.ai/docs/downloads/

Download OpenCV OpenVisionCapsules:

Read OpenCV OpenVisionCapsules Documentation:

openvisioncapsules.readthedocs.io/en/latest/

How to build OpenVisionCapsules compatible hardware | OpenCV Hardware Partnership Program

https://opencv.org/opencv-partnership-program//

Aotu Partner & support: https://forum.aotu.ai/

Email: info@aotu.ai


Smart machines give people the opportunity to improve the world around them. Aotu makes innovations that enable machines to see and understand, with the goal of being the world leader in rapid AI deployment for smart vision IoT and video AI customers.

1 Like