Adaptive Parking Slot Occupancy Detection Using Vision Transformer and LLIE

The number of vehicle users is increasing exponentially in this fast-growing economy, and so is the demand for parking space. We can utilize the intelligent parking system to avoid traffic congestion and save time for drivers. Various solutions have been developed in the past and the most prominent ones are using Convolution Neural Network on CCTV Camera. Object detection and instance segmentation are better than the classification methodology to detect parking slot availability. However, most of the papers lack identifying the parking slots in low light conditions, and they are specific to certain views such as aerial or standard. This research paper solved the above problems using the Low Light Image Enhancement technique and adaptive aerial/standard view by vision transformer and aerial view detector. However, there is no standard dataset to benchmark both the aerial/standard view parking slot detection in low light conditions. So we released a custom dataset containing annotated 1,257 images for public access.