In 2015 the the launch of YOLO — a high-performing computer vision model that could produce predictions for real-time object detection — sparked an avalanche of progress that accelerated computer vision’s leap from research to market.
Since then, it has been an exciting time for startups as entrepreneurs increasingly discover computer vision applications in everything from retail and agriculture to construction. With lower computing costs, greater model accuracy, and rapid distribution of raw data, more and more startups are turning to computer vision to find solutions to problems.
However, before starting to build AI systems, founders should think carefully about their risk appetite, data management practices, and strategies to future-proof their AI stack.
TechCrunch+ is holding a Memorial Day sale. You can temporarily save 50% on annual subscriptions.
Below are four factors founders should consider when deciding to build computer vision models.
Is deep learning the right tool to solve my problem?
It may sound silly, but the first question founders should ask themselves is whether they even need a deep learning approach to solve their problem.
During my time in finance, I’ve often seen that we would hire a new employee directly from college who would like to use the latest deep learning model to solve a problem. After working on the model for some time, they concluded that using a variant of linear regression worked better.
To avoid falling into the so-called prototype-production gap, founders need to think carefully about the performance characteristics required for model implementation.
The moral of the story?
Deep learning may sound like a futuristic solution, but in reality, these systems are sensitive to many small factors. Often you can use an already existing and simpler solution – such as a ‘classic’ algorithm – that produces an equally good or better result at a lower cost.
Consider the problem and solution from all angles before building a deep learning model.
Deep learning in general, and computer vision in particular, holds promise for creating new approaches to solving old problems. However, building these systems comes with an investment risk: you need machine learning engineers, lots of data and validation mechanisms to take these models into production and build a functioning AI system.
It is best to evaluate whether a simpler solution can solve your problem before embarking on such a large-scale effort.
Conduct a thorough risk assessment
Before building an AI system, founders need to consider their risk appetite, which means evaluating the risks that arise at both the application layer and the research and development stage.