The Eye of AI #5

Computer vision is a truly transformational technology for the retail industry, but it has applications that touch nearly every aspect of our lives. This is our weekly summary of what happens when cameras, computers and great ideas get together.

What's the Point of Mosquitoes Anyway?

We googled it - there really isn't any point to mosquitoes other than causing caribou to migrate across the tundra, and I'm pretty sure we could come up with a system of horns and flashing lights to do the same job. Other than that they are at best a low-key summer-spoiling nuisance and at worst a disease carrying menace. John Hopkins researchers have turned to computer vision to develop an app called VectorCam that can identify mosquito species quickly, helping target the exact culprits behind malaria's spread. This innovation empowers local health workers in places like Uganda to act faster and more efficiently, potentially saving countless lives by staying ahead of the disease. Now pass the deet.

The World's Fastet Jigsaw-Puzzling Robot

Ever start a puzzle and then simply give up and decide never, ever to use your dining table again? Well, help is at hand in the form of Jigsaw, the unimaginatively named robot that can solve a 1000-piece, all-white puzzle without human help. Built from a repurposed CNC router and equipped with a Google Pixel camera, Jigsaw meticulously photographs, computes, and assembles the puzzle pieces with the precision of a perfectionist. It took over a year of fine-tuning and programming, but Jigsaw’s success shows that even the most mind-numbing tasks can be mastered by machines—if you have the patience to teach them.

Smarter Urban Fire Fighting

As skyscrapers rise, so do fire risks, but Hong Kong Polytechnic University's SureFire system is here to outsmart the flames. Led by Prof. Usmani and Dr. Huang, this AI-powered innovation provides real-time fire risk assessment and forecasting, with 90% accuracy in predicting fire scenes minutes in advance. After five years of research, SureFire is transforming urban safety, guiding evacuations, and even empowering firefighting robots. Hong Kong is blazing a trail as a smart city leader, one high-rise at a time.

Stay In Your Lane

Staying in a lane ought to be table stakes for an autonomous vehicle, however the technology relies on certain things beyond its control like comprehensive road markings and anything that's not snow. Enter LD-CAM, a deep learning model designed to tackle such problems with precision. Using a mix of convolutional attention mechanisms, the model excels at recognizing lanes on curvy, damaged, or unmarked roads, achieving over 97% accuracy. While it outperforms existing methods, its complexity might strain low-power systems. Future improvements aim to reduce this overhead while maintaining high performance, making LD-CAM a promising tool for safer autonomous driving.

Simplifying Diagnosis of Neurological Conditions

Many degenerative neurological disorders such as Parkinson's Disease cause tremor symptoms in the patient, but the types of movement vary wildly. In a quest to simplify and speed up diagnosis, researchers have validated a computer vision algorithm, Mediapipe, for analyzing tremors in clinical videos, offering a simpler alternative to expensive motion capture systems. This AI tool tracks hand movements with impressive accuracy, even during complex tasks, and can predict outcomes of deep brain stimulation (DBS) for tremor disorders. Unlike traditional methods, this approach provides more detailed tremor insights, making it a game-changer in neurology for both diagnosis and treatment. It’s a promising step towards more accessible, precise tremor analysis in clinical settings.

That's everything for this week. Please keep an eye on the SAI Group blog for everything that we're talking about, including a new article this week about how false alerts from theft detection systems impact retailers.

Got some cool tech to share? Whether its your own project, or just something you saw and thought "I want people to know about this!", let us know about it and we'll include it in upcoming editions.