Computer Vision

Computer Vision
Computer vision is an interdisciplinary logical field that arrangements with how Computers can acquire undeniable level comprehension from advanced pictures or recordings. According to the viewpoint of designing, it looks to comprehend and computerize undertakings that the human visual framework can do.
Edge AI
AiFA makes superhuman recognition systems that disrupt existing processes.
AiFA’s software orchestrated, robust, edge-optimized Ai Camera systems accurately recognize objects and activities across many enterprise use cases to deliver insights and actions.
Edge AI has designed, prototyped, and produced in limited runs through manufacturing partners – the robust edge device, the camera module, illumination system, power supply system, peripherals, and firmware to control these items. Edge AI‘s optimized Ai algorithms have proven themselves in the real world. AiFA has run software on these devices in several use cases and have delivered near 100% accuracy in recognizing objects and activities. Edge AI’s Orchestration software efficiently manages users, devices, algorithms, artifacts, and actions.
With an API-first, tightly integrated offering, we make it easy to install, integrate, operate, scale and upgrade.
Edge AI - ViSRUPT

Edge AI - battle tested.
Digital Video Processing & Management
Developed digital video logistics, storage, play listing & display platform for advertising in cinemas and signages.
Acquired customers like IBM, TCS, Airtel, Kinetic, Reliance for digital video related software.
Ai Analytics & Autonomous Devices
Pivoted from digital video logistics to digital video analytics as market potential was enormous, demand is agnostic to cyclical variations, and the need is global.
Started building Autonomous Vision System hardware to meet the needs of high end vision Ai. Extended application to adjacent markets as they are severely under served.
With clients from USA, Japan and India – the company addresses the needs of high end, multispectral, mission critical autonomous vision requirements.
Integrated Devices - Sensors + Compute + Ai
EdgeAi Cam
EdgeAi Cam XT
EdgeAi Cam Ultra
+ Multiple sensors – video, audio mic, gps, motion
+ Illumination – IR, LED
+ Integrated power management, battery backup
+ On-board compute – integrated CPU, GPU
+ Ports for local connectivity to external devices
+ Ai inference models
+ Comprehensive software modules to manage sensor pipeline, alerts, media artifacts, API interface

Technology Capabilities
HW+Ai+SW
Artificial Intelligence
We have deep expertise in designing deep neural networks that can analyse videos, images,audio, sensor data to make sense of the real world.
Our artificial intelligence (Al) based algorithms achieve state of the art accuracy scores and are optimized to deliver inference in real time on edge devices.
Areas of specialisation:
- Deep NN, Convolution NN, Recurrent NN
- TensorFlow, Keras, MxNet, PyTorch, Caffe
- Video pipeline management -OpenCV, Media Foundation
- Platform optimisation – Intel OpenVino toolkit
Video HW & SW
We build device drivers to manage hardware modules for image processing, external interfaces, power etc.
We have in-house expertise to integrate diverse image sensors and image signal processors with our own video pipeline software to manage video frames, manipulate images formats and run video analytics.
Enterprise SW
We have experience in building enterprise grade software that
- ensures 99.999 level availability
- manages thousands of devices
- handles millions of digital media files – storage, archival, hyper-tagging
- provides robust security to secure data at rest and in flight
- easily integrates to other enterprise solutions through standard APIs
- is easy to maintain and upgrade
- is built on robust architecture principles for modularity and components, dockers
Areas of specialisation:
C++, Enterprise Java, Linux, Windows 10
Representative Solutions
Technology Capabilities
Strong Partnerships with Intel, Sony, Socio...
We are invited speakers in technology events to evangelise Ai and Edge Processing, Joint marketing collateral development (solution white papers + customer experience center)

Automotive Inspection
- Check homologation (Language, Correct Sticker)
- Check for presence or absence of connectors / bolts.
- Check for over machined bolts / screws
- Check for type of number plates used (based on model and delivery location)
- Check for scratches / dents on the vehicle body
- Check alignment of key components
Contactless Operations
- Confirm person is wearing mask and PPE
- Recognise people even when wearing mask
- Support contactless operations using gestures and audio inputs


Traffic Monitoring, License Plates, Container Numbers
- Real time analytics at the edge to detect and recognise license plates and container numbers
- Multiple cameras to read numbers on different sides of the container and reconcile the numbers to increase accuracy
Driver Monitoring, Vehicle Safety
Camera mounted on vehicle dashboard to monitor driver to detect fatigue (yawning, blinking, nodding) and distraction (looking away, using mobile)
Real time alerts served through sounds, audio messages, buzzer


Visitor Tracking, Audience Measurement
- Video analytics for a large consumer expo event
- 150,000+ visitors over 5 days
- Measured visitors, dwell and gaze time
- Determined which exhibits generated more interest for the visitors
- Helped the client test alternate models, assess
- Customer interest and launch the correct models for the market
Face Recognition, Analytics
Our state of the art face analytics engine delivers very accurate face recognition and analyses facial attributes to classify by age, gender, detection emotions, track gaze and head position, detect usage of mask, eyeglasses, helmet etc

ViSRUPT
Shaping modern work with superhuman recognition
Shaping modern work with superhuman recognition
The intersection of Ai based vision analytics and edge processing devices and software driven devices provides enterprises with new capabilities to enhance their operations in the field – serving customers in stores, service centers, quality inspection of products in plants and warehouses, managing safety and operations of vehicles, remote infrastructure etc.
By introducing the capability to detect, classify, recognize, track different objects and activities in real time 24 x 7, enterprises can drive significant value out of their current processes related to how they serve customers, how they enhance employee productivity, how they reduce leakages and defects in products and service delivery and how they ensure safety and good operations of their assets, premises, infra.
AiFA brings these key technology capabilities in a tightly integrated, resilient and easy to deploy and operate solution. Our solution makes it easy for enterprises to onboard these next gen technologies.
Technology innovations under the hood
Innovations
Use of raw signals for analytics – uncompressed, more accurate.
Al algorithms not just for analytics but for image processing to improve the signal capture system at source
Power management, mechanical design optimised for field performance
Flexible/ elastic both software and hardware easy to onboard new software algorithms as well as easy to integrate with external devices
See in the dark, Optical zoom with digital, removal of PTZ, Eventual removal of ISP layer, Fusion of multi-spectrum sensors, new optics, synchronised lighting etc. are some of the hidden advantages.
Solution Stack
Al algorithms-face recognition, Indian number plates, container numbers, object classification, motion analysis.
Pipeline management – in-device to manage different sensors
Dockers and containers for easy deployment and update
Video and media logistics encode, decode, encrypt, ship, archive, retrieve
Technology innovations under the hood
By working with many sensors and many algorithms, the 360° situation awareness allows for better inference and intelligent and instant actions. All the algorithms have been trained with own, generated and standard data sets. The fine tuning of the algorithms and its weights have allowed for better accuracy.
Finally all the algorithms have been optimized to run on the target architecture – thus ensuring speed and cost advantage.
Edge AI's Solution Stack
Ops command in Cloud
- Orchestrator
- Data Visualizer
- Remediator
- Knowledge repository
Ai at Edge
- Artifacts manager
- Action trigger
- Inference engines
- Optimized Ai algos
Device On Ground
- Purpose designed
- Resilient architecture
- Multiple actuators
- Multiple sensors
Process chain at a warehouse (with Ai models)

Our solution comes with readily available & minimal efforts Ai models for process chain at warehouse. Sensor fusion (RFID, Barcode, Weighing machine & Microphone integration) will cove many of the requirements
Flow of People
- People
- Vehicles
- Products
Stocking of
- Products
- Equipments
- Safety EQPMT