VisionBot: The Evolution of Visual Data

Amit Chakraborty,Co-FounderSince its inception, VisionBot has demonstrated exceptional expertise in object detection. In its quest, the firm has built core patented technology focused on custom object detection through training of neural network models. Visual inspection of objects in an enterprise workflow is also major source of error as it is still heavily dependent on manual reporting that shows inconsistency due to human fatigue. Thus, enterprises utilising VisionBot can drastically improve operational efficiency of their workflows while optimising costs. VisionBot platform's core competence lies in Automating visual monitoring. Given its ability to detect any type of object or event, enterprises which deal with a wide variety of objects and events can utilize its features in a plethora of functions. Some industries include Construction which uses a variety of vehicles and machineries, Logistics which deal in a large number of containers and shipments, Manufacturing which have large variety of inventory items, and Physical Retail that is required to track wide variety of items on display racks/ check out.

Imagine an enterprise with a large number of geographically dispersed factory and warehouse sites. While most of the operations in individual sites are connected over an ERP system, visual inspection is still manual and cannot be directly analysed by the computing system. These still depend on human operators keying in anecdotal data. While the enterprise already uses macro analytics tools for their other workflows they are unable to retrieve objective data from those processes that require visual inspection. This is where Visionbot platform offers the solution of obtaining objective inference from visuals captured through numerous cameras connected in an Enterprise network.

The Tech Involved
VisionBot is specifically utilizing Deep learning's Convlutional Neural networks (CNN) for object detection and Recurrent Neural Network (RNN) for sequential analysis that leads to event detection. The tech startup is implementing its custom version of generative Adversarial Networks (GAN) to optimize training processes."Our technology endeavour is currently focussed on optimising the neural network training process in terms of both duration and computing resources. There is a lot of
innovation happening in this space and we are building automation to help optimise the training itself", says Prashanth Rangaswamy, Co-founder, VisionBot. Rightly so, VisionBot's initial implementations were project based for large enterprises. However, the startup soon realized the challenges that project based implementations pose intermsof both scalability as well as customisation work impacting core development work. Therefore, VisionBot created a user friendly training module that end user teams can use with minimum learning curve. This has helped both, VisionBot's end users and system integrator partners, enabling them to provide custom solutions for themselves, and in turn proving Visionbot to be a truly exceptional Self service platform. Besides, platform allows customer teams to do the training without any dependency on VisionBot's tech team. This in turn helps VisionBot's team to concentrate on augmenting platform features and optimisations that keeps the team nimble and the operations to scale up.

Since its inception, visionbot has demonstrated exceptional expertise in object detection

Demonstrating Its Finesse In Real World Problems
VisionBot has been working with L&T constructions for automating their real time inventory control system from underground pipes transported from open yards to construction sites. The firm has also worked with Aditya Birla group companies for both, automation for their textile factory processes as well as their retail outlets. On the government side, VisionBot platform has been instrumental for KSRTC, wherein the firm automated the coordination between bus terminus and depot, thereby optimizing bus congestion at terminals. Overall, VisionBot has been working with more than 30 system integrators in India and around 10 overseas partners including UAE and the US. These partners practice their expertise in CCTV domain, thus giving VisionBot direct access to several end users within a short period of time.

Prashanth Rangaswamy, Co-Founder, Visionbot
Amit is positive that AI in computer vision is in nascent stage and is going to see phenomenal growth over the next 8-10 years. He says, "Most of the things that we do today will be done remotely through real time camera feeds. It's only AI that can derive data from such humongous number of cameras in daily lives". Clearly, computer vision will play its part in almost every facet of our work in homes, office and factories. Given the stellar work that it has been doing, Visionbot with its cloud first platform, is well positioned to leverage the growth and achieve market leadership in this burgeoning market.