TRODO: a public vehicle odometers dataset for computer vision

The TRODO dataset is a collection of annotated odometer images from several vehicles operating in a real-world product delivery network. The dataset contains a total number of 2,389 raw images with proper annotations. Furthermore, TRODO also includes the ground truth information of odometer types and mileages of vehicles. In the fields of insurance and logistics, the TRODO dataset can be utilized to build smart vision systems for mileage reading from odometer images. It is also possible to extend the application areas of the dataset by transfer learning. Researchers can use the dataset to train, validate, and test novel or existing supervised learning approaches.