To determine the number of available spaces to park a vehicle from a given image of a parking lot.
The file map.csv can be created by the python script in map_creator folder. Use the following command to execute:
python map_creator.py
This opens up a GUI which you can use to mark the spaces which are parking lots. This map can also be automatically generated using object trackers and dwell time. We will integrate that part with this code soon.
You can run the light weight object detector by downloading the pretrained weights for mobile ssd. This file can be found at:
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/detection_model_zoo.md
Place the frozen_inference_graph.pb file for MobileSSD trained on COCO Dataset in the models folder.
python3 main.py src/img/input.jpg dst/img/output.jpg
The output in the terminal shows the number of available parking slots by matching the processed image map and the map created by the user using the GUI. The operation takes the average area of bounding boxes of the vehicles to compute the number of unoccupied slots. In addition to this, we also show the predicted frames per second the algorithm is expected to perform.