Master's thesis by Mateus Silva in the postgraduate program in electrical engineering at UNIFEI Itabira campus.
Topological localization system for smart vehicles using a two-stage deep learning pipeline.
The system identifies the vehicle position in a map of predefined nodes:
- Model 1: detects
straightvsintersection - Model 2: classifies the intersection into one of the nodes (16 classes)
- main.py — application entry point.
- gui.py — demonstration interface.
- src/ — main package with modules: capture, gps, gui, Models, Navegation, tracker, utils.
- src/Models/ — trained weights and modeling scripts.
- data/ — logs and images used for training/testing.
- maps/ — map configuration files.
- Validation/ — logs, mages, and videos used for field validation of the complete pipeline.
- notebooks — experiments and analyses (see the .ipynb files listed above).
- Create and activate a Python 3.10 environment (recommended).
- Install dependencies: pip install -r requirements.txt
- Graphical interface: run gui.py.
- To reproduce experiments and train models, refer to the notebooks: Model_training_1.ipynb and Model_training_2.ipynb. The model-building function is in
build_model.
Mateus Silva
MSc Student in Electrical Engineering – UNIFEI
Automation & Control Engineer
Phone: +55 31 9 8818-8696
E-mails: mateusfilipi22@unifei.edu.br | mateus.filipe.22@outlook.com