Self-Driving Cars Struggle to Detect Children And Dark-Skinned Pedestrians: Study

A new study has found that self-driving cars may not be the most reliable when it comes to detection of pedestrians. For perspective, self-driving cars are driven by artificial intelligence (AI) algorithms trained on large models of data. What these cars do on the streets represents these algorithms on which these vehicles were trained and built.

According to a new study, AI used for self-driving vehicles is less likely to detect children and dark-skinned pedestrians. The study tested eight different pedestrian detection software wherein they were shown 8,000 images of pedestrians from different age groups, both women and men, and with different skin tones.



AI systems, in this study, struggled to recognise human faces of children and adults with dark skin tones. The same systems were 20% more accurate with adult faces, and 7.5% more accurate detecting pedestrians with lighter skin.

Also read: Former Apple Employee Flees To China After Allegedly Stealing Self-Driving Car Data

What this essentially means that the AI software running these vehicles need more training before they're truly read to take to the streets. AI learns about the faces through images and data fed to its model from different angles.


Also read: Problems In Tesla's Self-Driving Tech Force Company To Recall 360,000 EVs


The AI systems were also unable to detect adult and children faces when lighting was low. This puts everyone at risk of collision with a self-driving car at night when the light is naturally low.


What do you think - are self-driving cars ready for the streets? Let us know in the comments below. For more in the world of technologyandscience, keep readingIndiatimes.com.