Challenge

AISG–SLA Visual Localisation Challenge

Develop AI models to accurately extract camera pose data from 2D monocular images.

USD 35,000 awarded
jul 2023
335 participants

Frequently Asked Questions (FAQ)

1. What is the camera coordinate system and the rotation order used?

Please see this official page from the manufacturer.

2. What are the X and Y axes?

For the raw images, the long side is the x-axis, and the short side is the y-axis. Note that the example images on the Problem Description page have been rotated for presentation purposes only.

3. What do the Easting, Northing, and Height mean?

The coordinate system we used is SVY21 (instead of the common WGS84). In this plane coordinate system, easting increases from west to east, northing increases from south to north, and elevation increases from down to up.

4. Should we return the world coordinates, as given in the training set?

There is no need to return world coordinates.

5. When designing and evaluating my model using the test data, how can I determine the origin?

Because the performance metrics used for evaluation are relative pose errors, you may use any 6-tuple as the origin, such as (0, 0, 0, 0, 0, 0) or (123, -56, 32.1, -56.78, 40.79, 4.56).

6. How to convert the “Easting, Northing, Height, Roll, Pitch, Yaw” into a 4x4 matrix?

See the functions euler_to_quaternion and transformation_matrix from the evaluation script normalised_rpe.py located on the data download page. First, use euler_to_quaternion to convert Euler angles (roll, pitch, yaw) to quaternion representations. Then, use transformation_matrix to convert the translations (easting, northing, height) and the quaternions to 4x4 matrices. You can see examples of these functions being used in the normalised_relative_pose_errors function in that script.

7. Can you provide an example on how to generate a submission?

We have prepared a script to demonstrate how a submission can be generated based on a pre-trained model. The script can be downloaded from the data download page—see DROID-SLAM-example.zip. Please note that this is only an example.