Estimate object trajectories in an automotive context, using KITTI dataset, with Mask-RCNN.
A very detailed description of the implamentation is available in the relazione.pdf
file (but in italian).
Three python scripts, each with different functions:
- Kitti-dets : Calcute tracks with predicted boxes and accuracy with motmetrics library.
- Kitti-mask : Calcute tracks with masks.
- kitti-features : Calcute tracks with features vectors.
User opens a selected script and leaves hashtag from main in final line of code and can insert the parameters of IouTheshold, ScoreTreshold and NoMatchAllowed from console.
The implementation has been done with the programming language Python 3.Project dependencies are the following Python modules.
Requirement | Version Used |
---|---|
Python | 3.6 |
matplotlib | 2.2.2 |
numpy | 1.14.5 |
motmetrics | 1.1.3 |
opencv-python | 3.4.3.18 |
The Mask-RCNN code is based on the following GitHub repository: