UBCSailBot-VideoTrainingData is a tool intended to ease the creation of training data for machine learning from human input. This tool converts the human input to generate a file with .label extension that is used as training data for obstacle detection algorithm. The standard format for this file can be found at demo_metainfo_label.
Install the requirements and opencv
sudo pip install -r requirements.txt
Run the script and pass in a video file as the first argument
python main.py boat.avi
The control scheme is the following:
| Button | Action |
|---|---|
| d | next |
| a | previous |
| w | skip 24 frames back |
| s | skip 24 frames ahead |
| n | mark frame as empty |
| z | mark frame as uncertain |
| esc | quit |