Hundreds of thousands of videos are uploaded to sites like YouTube, Facebook, and Instagram every minute. Analysing the content of reams and reams of videos is no small task. Netra, started in 2013, provides AI tools to understand and organise video content. “Video analysis lacks proper toolkits. Additionally, video is informationally the richest medium. There is a lot happening on screen. It’s not just about understanding objects in there, but also about understanding the context around those objects. Video is orders of magnitude richer in information density than text or other data types. It lies largely unexplored with the current paradigms. We built our technology, to address this challenge,” said Shashi Kant, cofounder &CTO of Netra.
“Our very first investor was Mark Cuban (of Shark Tank fame) who responded to a cold email way back in 2015, at a time when the current AI Spring was just breaking out. He and a few other leading investors could see the approaching AI revolution ahead of many others. Mark Cuban participated in the original seed round, and has invested in our follow-up seed rounds,” Shashi added.
Netra was built on top of its founder’s research at MIT CSAIL. “Video is a different beast due to its sheer size and complexity. We had to make some significant technical breakthroughs to support video at an affordable price without compromising on accuracy or features. Squaring that circle has been our technological edge. Ultimately we like to differentiate ourselves by delivering immediate and significant ROI to customers,” he said. The company’s algorithm can recognise actions, objects, emotions, places, etc in a footage.
“We provide a way to support custom domain-specific taxonomies. We have learned that even though our partners want to start with context and safety, eventually it is the search and taxonomy creation that provides the ultimate value for monetising their video assets,” he said.
The company uses a variety of approaches to process video, including some deep learning techniques, combined with traditional machine-vision algorithms.
- GPUs for processing videos
- FFmpeg, VLC and OpenCV for processing videos
“We do not require any audience identification features provided within the metadata. Our technology mainly understands the content without requiring any Personally Identifiable Information (PII). In addition, we have ensured that we do not use any facial recognition or biometrics within our offering. This ensures that our partners do not have to worry about privacy,” Shashi Kant said.
Netra is assisting corporations in pairing videos with appropriate advertisements to avoid tracking users.
“We started in the media and entertainment verticals, which still is the highest revenue generator for us. We work with video platforms, broadcasters and other data platforms. But additionally we have grown to work with martech players who want to categorize and classify their video assets in order to provide insights to their clients. Also manufacturing and supply chain, where they are amassing video footage from manufacturing facilities whereby we can help them solve operation efficiency issues by identifying people and vehicle movements. We also work with information monitoring businesses that track real-time video feeds to detect anomalies, disruptions, etc. to provide this information to news channels, financial firms and law enforcement organizations,” Shashi said.
“We compete directly with the video analysis APIs provided by Google Cloud Vision, Amazon AWS, Microsoft Azure and some other lesser-known businesses. But our niche is focusing on video and offering a pizza-and-toppings business model whereby users can pick and choose whichever combination of analytics they need for their business needs. For example, we offer Context and Safety tags as standard offering, but customers can choose: video activities (e.g Surfing, working etc), places (e.g. home, playground, beach etc), affinities (e.g travel enthusiasts, foodies etc), and a bunch of other specific models that are designed to assist businesses extract maximum value from their video assets,” added Shashi.
The company has received a seed funding of $4 million from investors like Mark Cuban, NXT ventures etc.