This project leverages computer vision techniques, streamlit for visualization, and an object classification model to perform real-time classification of parking spaces as either empty or occupied. The main objective is to provide an efficient and accurate system for monitoring parking spaces using drone footage.
The project utilizes a dataset that can be accessed from below url. The dataset includes annotated images of parking spaces, indicating whether each space is empty or occupied. https://drive.google.com/drive/folders/1CjEFWihRqTLNUnYRwHXxGAVwSXF2k8QC
The project includes a model training code that enables users to either use a pre-trained model provided in the project or train their own model. The model is designed to perform robust classification of parking spaces in real-time.
Real-time parking space classification using computer vision. Streamlit for interactive visualization and user interface. Utilization of an object classification model for accurate predictions. Efficient model training code for flexibility in model selection.
The upcoming enhancement aims to incorporate the relative positioning of bounding boxes concerning the drone's location. This improvement enhances the prediction model's accuracy and generalization in real-world scenarios. By considering the spatial relationship between parking spaces and the drone's position, the system can provide more context-aware and reliable predictions.