The Eye of AI #4

The Eye of AI #4

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.

Stress-Free Weighing

What does a mouse weigh? Photo: ScienceBlog

Wouldn't it be great if getting on the bathroom scales didn't involve a certain amount of fear and trepidation? Well, this isn't quite like that. Unless you're a laboratory mouse. Apparently the stress involved in weighing a mouse to ensure the proper test dosage of an emerging treatment can cause a blip in the dependability of the results. With this in mind, a team led by Jackson Laboratory has developed a method to calculate the mass of an individual mouse without the need to pop them on the scales.

All Aboard the Safety Ship

Collision avoidance at sea. Photo: FredOlsenExpress

Preventing collisions at sea is a complex task. Over hundreds of years a web of technologies and processes has evolved including having people on watch 24 hours a day, lights of different colours, buoys, horns, lighthouses, and laws for which side ships can pass each other. Even with radar and position tracking systems, accidents still happen and cost hundreds of lives. Ferry builder Fred Olsen is starting to install computer vision systems on their ships to detect things that current systems just cannot see - unsignalled craft, floating obstacles, inflatables, kayaks and persons over board. That's something we can all get onboard with.

Bovine Behaviour Bots

Just cows doing cow things. Photo: Inner Mongolia University

Cows are not complex creatures. They eat. They walk around. They drink water. When it's about to rain, they lie down (apparently). From a farming perspective though it is important to know what is happening in your herd. Are they drinking enough water? Are they spending too much time lying down? Could this behaviour be an early indicator of illness or disease? Researchers at Inner Mongolia University in China, where cattle populations are growing quickly to meet local demands for beef and dairy, have just released a new dataset for anyone hoping to develop their own herd-management systems. While the current dataset is good for broad analysis of herds, leveraging YOLO+DeepSort could assign unique IDs to each animal for individualised reporting.

It's not a Bug, It's a Feature

Automatic insect classification. Photo: MIT

Insects are a vital component of the world's ecosystems, acting as pollinators and nutrient cyclers, as well as propping up the food chain for birds, small animals and even humans. Their numbers are in drastic decline however, with 65% of species facing extinction this century. Researchers at the Massachusetts Institute of Technology have turned their attention to rapid classification and identification of insect species by scraping the web for 1.54 million images which they used to train a custom YOLOv8 model. The results are almost 90% accurate in identifying 2,500 species, giving biologists and climate scientists greater insight into the changing environment.

Analysing ASD Through Body Movements

Students analysing body movements to identify ASD traits. Photo: Indiana University

Non-verbal communication in humans permeates nearly everything we do and makes us the social creatures that we are. We pick up on the smallest of cues from people around us, often without even realising. Unless you have Autism Spectrum Disorder (ASD), that is. ASD closes off the brain's social pathways making it difficult to relate to others, including by picking up on the cues the rest of us take for granted. Without an understanding of the minute gestures that neurotypical people give off, it is extremely difficult for those with ASD to replicate them. At Indiana University a collaborative effort between the Kinesiology Department, the Intelligent Systems Engineering team, and the School of Informatics, Computing, and Engineering aims to detect ASD traits in otherwise neurotypical people. The hope is to develop minimally invasive diagnostic tools, with potential patients able to upload video of themselves rather than attend clinics and have sensors placed on them.


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 combatting the rise in retail shrinkage.

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.


要查看或添加评论,请登录

SAI的更多文章

社区洞察

其他会员也浏览了