Software is eating the world, and AI is eating software. Revolution in Video Analytics is no different. Traditionally, CCTVs were being used for live surveillance by a security team or for post-event analysis.
But there are three broad technologies which are converging and unlocking new humongous potential:
Pervasive cameras have made it possible to record every human action and monitor every process. It used to be difficult to store and transfer large files - but ever-increasing Internet bandwidth along with low cost cloud infra have resolved this to a large extent. The biggest revolution however is the advancements in deep learning algorithms enabling analysing videos. Now, virtually any process can be measured and monitored - be it human or a machine.
These trends have opened up entirely new dimension of use cases - which different industry verticals are still exploring and experimenting with. These are wide-ranging but let me discuss three major emerging Industrial use cases:
Industries still employ large workforce. Supervising such a large work force becomes a challenge in itself and companies are now adopting Video Analytics to help supervisors and managers keep track of the work in the factory floor. Videos can track how much time each worker spends in each process and which tasks slow the production line.
Video analytics is also evolving as an alternate to IoT devices and sensors as it can efficiently do the same process monitoring cheaply and effectively. Auto ancillaries, textile and other such industries with large worker footprint are now adopting these approaches with great promise.
Quality Checking and Vision based sorting
Using Vision for quality check on a production line is not a very novel idea. Even few decades ago, expensive industrial cameras clicked precise images and matched pixel to pixel to check for product defects. AI algorithms now enable regular vision camera to achieve the same performance and at the same time are robust to changes in environmental conditions.
Foxconn with help of Landing.ai is deploying such algorithms for measuring defects in mobile phone chips. These solutions are now fast being adopted by FMCG, footwear, steel and food processing industries for identifying defects on production lines. At Silversparro, we have also worked for automating metal sorting for a large US player.
The use case which we care deeply about is the improvement in worker safety. A simplest use case is using AI to detect if workers are wearing Hard Hats and other suited protective gear. This can prevent several injuries and accidents. More complex cases involve raising alarm if workers seem to be entering an accident-prone area or are in proximity of equipment with extremely high temperatures.
A chemical plant working with dangerous fumes reached out to us to detect if workers become unconscious while working. Another requirement was to raise an alarm if a worker lowers himself in a well or a pipe and does not return within 30 minutes.
We are confident that AI powered Video analytics will become standard solution for worker safety.
There are still some challenges that need to be resolved. Dynamic and noisy industrial environments can lead to poor accuracy of the models. Hiring engineers with AI skillset to deploy and constantly improve accuracy of models is still a challenge. The norms for ensuring privacy of workers are still evolving and it remains an open issue for now.
The race to adopt and extract value from AI powered Video Analytics has begun and we are excited to see how it unfolds.
Silversparro provides AI-powered Video Analytics for workplace productivity. Silversparro is founded by IIT Delhi Alumni – Abhinav Kumar Gupta, Ankit Agarwal and Ravikant Bhargava and is working for clients such as Viacom18, Policybazaar, Aditya Birla Finance Limited, UHV Technologies etc. Silversparro is backed by NVIDIA Inception program and marquee investors such as Anand Chandrashekaran (Facebook), Dinesh Agarwal (Indiamart), Rajesh Sawheny (Innerchef) etc.
Posted at YourStory.