- Clone repo & cd to it
cd convenient-parking/
- Create venv & activate
python3 -m venv /path/to/new/virtual/environment
source /path/to/new/virtual/environment/bin/activate
- Install requirements
pip3 install -r requirements.txt
- Install weights file from M-RCNN
git clone https://github.com/matterport/Mask_RCNN/releases/download/v2.1/mask_rcnn_balloon.h5
- Run the program
python parking.py
- Don't forget to deactivate venv Checkout to the Dockerized branch if there are any problems with requirements.
An application for recognizing free parking spaces. Mask R-CNN (model for object detection and instance segmentation on Keras and TensorFlow) was used in the implementation. I recommend to read its documentation, download weights file and move it to the project folder. No manual segmentation of the parking lot is required to detect available parking spaces, the space will be marked as free after the car mask disappears from there. Due to certain difficulties when running the program on different devices, it was decided to make a Dockerized version (located in a separate branch).
I hope i will add the version that works with 2nd TF, usage of TF v.1 causes too much warnings.
That's about all...