I estimated an AV's trajectory by images taken with a monocular camera set up on the vehicle.
In particular, I did
- Extract features from photos
- Use the extracted features to find matches between the features
- Use the found matches to estimate the camera motion between subsequent photos
- Use the estimated camera motion to build the vehicle trajectory
See the Jupyter Notebook for more details.
Given the output from
- semantic segmentation neural networks
- 2D object detectors
I did
- Implement drivable space estimation in 3D
- Implement lane estimation
- Filter errors in the output of 2D object detectors
- Use the filtered 2D object detection results to determine how far obstacles are from the AV
See the Jupyter Notebook for more details.