Artificial Intelligence Smart Eyeglasses for the Detection and Description of Stationary Objects
Robert J. Medina, Salaam Botros, Pranali Gandhi, Rehan Choudhury, Kalpita Das, Antonio Bechara Ghobril, Carol L. ShieldsImportance
Artificial intelligence (AI) smart eyeglasses may have potential uses for patients with low or no vision, but evidence is needed to understand potential benefits and limitations of their use.
Objective
To evaluate the accuracy of AI smart eyeglasses (Ray-Ban Meta AI eyeglasses [Generation 2]) on single and multiple object identification and description.
Design, Setting, and Participants
This case series was a pilot feasibility study involving 6 study authors as the participants. A white tabletop and background were used for all tasks unless otherwise specified.
Main Outcomes and Measures
The primary outcome measure was AI model accuracy defined as percentage of correct responses for single object identification, color discrimination, directionality, big and small object counting, reading (medication labels, food labels, handwriting, children’s books), and US paper and coin money identification and counting tasks.
Results
The participants included 3 males (50%) and 3 females (50%); their mean age was 30 years (median, 29 years); mean height, 67 inches (range, 61-74 inches); and mean age at English acquisition, 3 years (median, 3 years). The smart eyeglasses identified common objects with 99% accuracy (699/700 trials; mean accuracy, 99%; 95% CI, 97%-100%). On object description tasks, color discrimination accuracy was 64% (286/450; mean accuracy, 62%; 95% CI, 51%-73%), object directionality was 83% (249/300 trials; mean accuracy, 81%; 95% CI, 72%-90%), and object counting was 50% (199/400 trials; mean accuracy, 49%; 95% CI, 40%-59%). For reading, standard text accuracy was 59% (59/100; mean accuracy, 59%; 95% CI, 45%-74%), for handwriting, mean accuracy was 88% (median, 93%; mean participant accuracy, 88%; 95% CI, 76%-99%), and for children’s books, mean accuracy was 93% (median, 100%; mean participant accuracy, 81%; 95% CI, 60%-100%). Individual money identification accuracy was 91% for paper (181/200; mean accuracy, 91%; 95% CI, 85%-97%) and 2% for coins (3/150; mean accuracy, 3%; 95% CI, 0%-8%).
Conclusions and Relevance
AI smart eyeglasses may offer a unique intervention for patients with low to no vision, performing best for identifying common objects, identifying neutral colors, and reading children’s books. AI smart eyeglass users should be aware of current limitations, which might improve as technology evolves in this field. Further studies are needed to define these benefits and limitations with patients who have low to no vision.