Every vehicle identification system must be evaluated on a dataset to be compared with other systems. So, the availability of an appropriate vehicle dataset is important for such systems. A standard dataset must accommodate different environmental conditions. The more variations in the dataset make it more challenging.
Our dataset, called IRVD , developed in different illumination and air conditions including rainy weather, sunny, cloudy, moonlit night, and dark night. The night images are also taken in two ways, one using an LED projector and the other without a projector. The cameras were installed at different heights from 0.5m to 7m. After capturing videos, the frames containing vehicles are cropped using a car detection algorithm that includes the entire car frame. Finally, the images are normalized to 256 × ۲۵۶ pixels.
The following file contains 10% of the whole datset. To get the complete version of the dataset, contact us. Please write why and for what purpose you want to use the dataset.