Our goal is to build a geolocator that can recognize license plates in images/videos in order to aid law enforcement in locating persons of interest. This is done using two separate machine learning models sequentially:
- The first model detects licence plates and isolates them from images/videos (adapted from https://github.com/theAIGuysCode/tensorflow-yolov4-tflite/tree/master)
- The second model takes the isolated images from the first model and predicts the top 5 most probable states that a plate is from (using ResNet: https://github.com/tornadomeet/ResNet/tree/master)
Scripts for training and setting up each of the models can be found in the branch labeled "ml"
- Clone the repository
- Download the model files from the Google Drive from the repository and save them in the model folder.
- Run pip3 install -r requirements.txt. This installs all the necessary libraries.
- Run python3 app.py
- Open http://127.0.0.1:5000/ on any browser to view the UI.
Google Drive link : https://drive.google.com/drive/folders/1XJRzu6WLNre8euyKdlRqHElFOu0mZO_t?usp=sharing
- The Detector model is fully implemented according to our specific needs and incorporated with the most recent State Classifier model
- The State Classifier model works with high accuracy on current test data
- Frontend is fully implemented. Upload, results pages works as intented
- Backend operations include server-side tasks, data processing, and interfacing with ML models
- Frontend provides the user interface for interacting with the application
- Have a profile page (multi user application)
- Keep track of User history and have a specific profile page=
- Integrate backend with Immich
- Find more data for training the State Classifier, particularly videos to be run through the first model, as well as realistically bad quality data