Safe and fair AI-based drug detection

Purpose and goal:

The purpose of the project was to collect data on the eyes of people affected by drugs and, with the help of developed algorithms and AI, find a method to identify drug effects in the eye area.

The goal was to develop a method that, via screening of the eye area, can ensure drug exposure with high accuracy. The method could be applied in a software for, among other things, vehicles to reduce drunk driving and increase traffic safety. Expected results and effects:

In dialogue with the market it emerged that the need to detect alcohol influence was priority 1. For this reason, we started collecting large volumes of data on people under the influence of alcohol in parallel. Although the model will continue to be improved, we have succeeded in developing an accurate algorithm that detects the influence of alcohol via a short film sequence (about 87% accuracy with a blood alcohol level of 0.5).

We currently have collaborative projects within the automotive industry with the goal of integrating the software into vehicles to improve traffic safety. Approach and implementation:

In collaboration with hospitals, we have collected data on people affected by drugs. In connection with that these people submit a drug test at the treatment clinics, filming of the eye area is also carried out. The filming has subsequently been linked with the results of the drug test.

The drug test has consisted of either a urine, saliva or blood test. The videos of people under the influence of alcohol have been collected at various events where people drank alcohol and were then filmed at different blood alcohol levels. These films have been the basis for developing the algorithm.