The sustainability goals of the aviation sector demand innovative ways to reduce CO2. One of the routes is improving the alignment of fuel, speed, and flying height to the weight of the airplane. The more accurately the weight is predicted, the better all parameters can be aligned. Passengers’ weight is the only weight variable not being measured but estimated by calculating the sum of the average weight of all passengers on board. This means the biggest improvement can be obtained by reducing the inaccuracy of the passengers' weight estimate. In line with this, Bit teamed up with a multinational airline company to improve the prediction of passengers’ weight without disturbing the boarding process or violating individual privacy.
To find out someone’s weight without using a weighing scale, we assumed the best route would be to predict a person’s body volume with a depth sensor. Research has shown that the use of this technology enabled the predictions of one’s BMI by analyzing a face, and our reason to believe there would be equal possibilities in the area of weight prediction. Having laid the ground for the prototyping phase by collecting weight and visual data at the airport, our core challenge appeared to be the transformation of the camera images into data to train the AI model. As such, we created features based on the images, such as height, age, and volume, followed by continuous testing and iterations until our model could manage to predict weights with minimal deviations.
people voluntarily recorded
frames of data
hours of data collection
Our team designed and trained an AI model that reduced weight deviation by 88%. Furthermore, the prototype shows there is potential in using computer vision to improve weight calculation, leading to reduced CO2 emission and six-figure business cost savings.
of the weight deviation in the AI model
Computer vision can be used for weight calculation
Estimation of six figure business cost savings