Master's thesis · Computer Vision
Recognizing places with a phone camera
A PyTorch pipeline that recognizes tourist landmarks in real time by matching the camera image against a database of descriptors. Exported to ONNX, with a sample Android app.
- Role
- Design and development (Master's in AI, VIU)
- Period
- 2025
The idea
Point your phone at a building or monument and instantly know what it is, without relying on precise GPS or on the place being tagged.
How it works
- Detection and descriptor extraction from the camera image with PyTorch models.
- Retrieval against a database of known-place descriptors to find the closest match.
- ONNX export to run the model on the device.
- Sample Android app that shows the result with augmented-reality anchors.