This week in AI & Machine Learning: A new object detection method, AI for Arbor Day, and a predictive model for healthcare providers.
A Note from the Author
You probably don’t associate Arbor Day with many time-honored traditions, but it’s the perfect time to reflect on wildfires and the threat they present to both local communities and the environment as a whole. Typical approaches to monitoring and preventing wildfires are reliant on humans to recognize subtle signals and take quick action. Vision AI can help turn visual data like drone and satellite imagery into an even more valuable source of data by training predictive fire-prevention models. Learn more about AI’s role in fighting wildfires in California and all over the world.
Custom vision AI models can support wildfire prevention, monitoring, and response efforts.
Artificial Intelligence News
A New Approach to Object Detection
Object detection is a fundamental computer vision technique with which a model identifies and localizes all of the relevant objects within a given image. A pair of Google researchers recently unveiled Pix2Seq, an alternative to popular object detection methods like Fast R-CNN, YOLO, and DETR. Simpler than those methods, Pix2Seq teaches models by describing images with a series of tokens pertaining to class labels and the coordinates of each bounding box. The researchers put their approach to the test on the popular COCO object detection dataset and achieved results similar to those achieved with Faster R-CNN and DETR.
Understanding the Differences Between AI and Doctors
One of the primary challenges for introducing AI into healthcare settings is the medical community’s general lack of insight into how and why AI solutions make the decisions they do. While it’s clear that AI could offer many benefits to doctors and patients, it’s decidedly less clear how humans and technology can best work as a team. Researchers from the University of Warsaw and New York University have assessed the differences in how radiologists and deep neural networks (DNNs) analyze breast cancer screenings. Doctors and AI, they found, took vastly different approaches in diagnosing malignant tissue. AI often reached accurate diagnoses by focusing on tiny, scattered details that doctors ignored while disregarding some more obvious indicators. The researchers hope their efforts can help others better understand both the capabilities and limitations of AI for diagnosing medical conditions.
Could AI Help Cut Down on Missed Appointments?
Researchers at Boston Children’s Hospital (BCH), where 1 in 5 patients is a no-show, have analyzed over 150,000 appointment records to develop a machine learning model capable of predicting when a patient is likely to miss an appointment. Currently, many health systems contend with no-shows by overbooking appointment slots. Things can go smoothly when hospitals and practices overbook, but underserved patients can just as often wind up facing even more obstacles to care. BCH is exploring a number of alternatives to overscheduling, including this new model. Questions remain as to how the model could best be put to use. Kathleen Conroy, co-author of the study and a clinical chief at BCH, notes that the hospital will need to exercise caution in even discussing the model and how it works. “We need to be deeply cautious . . . so we don’t accidentally create bias amongst our staff around patients who they might be reaching out to.” Read up on more of the ways AI is supporting healthcare providers everywhere.
Join our Community
See you next week! Until then, keep the conversation going on Plainsight’s AI for All Slack Channel.
About the Author & Plainsight
Bennett Glace is a B2B technology content writer and cinephile from Philadelphia. He helps Plainsight in its mission to make vision AI accessible to entire enterprise teams.
Plainsight’s vision AI platform streamlines and optimizes the full computer vision lifecycle. From data annotation through deployment, customers can quickly create and successfully operationalize their own vision AI applications to solve highly diverse business challenges.
