Appearance-Based Teach-and-Repeat Navigation Testing
| Data Description |
Dataset inlcudes 3 image sequences. Each sequence is composed of multiple image sets taken from the subsequent positions on a testing path. Each set contains 9 images taken with various camera shifts and rotations. As the dataset is designed for appearance-based teach-and-repeat methods, the transformations are lateral translation and horizontal camera rotation, simulating various differences from the original trajectory during repeating.
Note: An example Python script to process the data and test your methods is included.
Image sequences:
- "Outdoors_day_winter": Day-time sequence taken in a university campus, combining urban and natural environments.
- "Outdoors_night_winter": Night-time sequence in a university campus, combining urban and natural environments.
- "Indoors": Indoor sequence taken in corridors of a university building
Note: Sequences 1 and 2 can be combined for the testing since the images were captured from equal positions.
| When using the dataset, please cite: |
- It is going to be added later...
| Contacts: |
Tomáš Pivoňka: tomas.pivonka@cvut.cz, Václav Truhlařík: truhlvac@cvut.cz