See through Computer Vision

Reading Time: 3 minutes

AI owes it big time to 2012. That year, for the first time, a deep learning neural network won the ImageNet classification challenge, beating the next best by 10.8 percentage points. Training deep networks at scale was impractical until then; the breakthrough innovation of training on Graphics Processing Units (GPUs) triggered a renaissance that led to a steady decline in error rates in the annual ImageNet challenge. The following year, another submission based on a neural network algorithm won the challenge. By 2014, all the high-scoring competitors in the competition had embraced neural networks.

The continuous stream of both positive and negative buzz surrounding emerging technologies often pushes organizations to make resource allocation decisions that are not well aligned with their interests. From embracing risky innovations without fully understanding their potential value to overlooking valuable-yet-less visible opportunities, most enterprises fluctuate between these extremes.

The Gartner Hype Cycle is a visual guide informing a technology’s potential impact on lives, businesses, jobs, and society in general. Characterizing the progression of concepts, technologies, and platforms into five distinct stages of development helps business leaders, technologists, and investors contextualize the developments as the ecosystem evolves.

In the inaugural edition of Hype Cycle for Artificial Intelligence, 2017, Computer Vision, a field based heavily on neural networks, debuted just past the Peak of Inflated Expectations. 

Surviving the Trough of Disillusionment and the Slope of Enlightenment, Computer Vision is now at the cusp of a breakout moment, entering the Plateau of Productivity in 2023. 

The promised land

Entry into the Plateau of Productivity demonstrates the robustness of a technology and its adoption by the mainstream. In recent years, Speech recognition, Virtual Reality and blockchain were successful incumbents on the plateau and now have well-entrenched use cases across a wide range of industries. 

Per IDC, the total worldwide market for computer vision technologies will grow to more than $2.1 billion in 2023, from $760 million in 2020, with a compounded annual growth rate of 57% expected through 2025, to a total market value of $7.2 billion. Augmented Reality to autonomous driving, robotics to retail, security & surveillance to smart agriculture, mainstream adoption of Computer Vision has today impacted domains as diverse as consumer electronics, healthcare, manufacturing, medical imaging, satellite imaging and spotting stock outages in stores. Extracting attributes from product images, tagging and deploying the components to custom-generate PDPs and taxonomies; Intelligent Document Processing of any type of record through astute extraction, multi-way matching, and reconciliation; instantaneous image moderation that validates image quality and appropriateness of content; extraction and enrichment of data from text, images, videos and more, either for use across business systems or to build and deploy models on top of it – the use cases for Computer Vision are plenty.

Since 2019, Vue.ai has deployed Computer Vision-based AI solutions for clients in NAM, EMEA and APAC markets. Vue.ai’s standards-setting work in this space has been recognized every year in the Gartner® Hype Cycle™ for Digital Commerce, for Visual Search since 2021.

Although the domain of neural networks has grown exponentially over the past decade, it is still perceived as a black box, whose failure modes can’t be reliably predicted or understood. Add up the five-nines reliability (99.999%) mandated by manufacturing and engineering use cases, Computer Vision’s fault lines stand exposed. Our real work begins now.

Read More About Intelligent Document Processing Article