You are about to access the Test Set or related features.
This dataset is part of the AI Cup 2026. By proceeding, you agree that:
This is an interactive visualization and Feature Engineering tool for the AI Cup 2026: Bird Track Radar Challenge. It allows you to explore bird flight paths and debug AI models directly in your browser. Github Repository
You can write Python code in the left sidebar to extract features or filter the dataset in real-time.
airspeed, radar_bird_size, plus
injected debug columns. meta['is_test'] is True
for test set tracks.
Your function must be named calculate and return a flat Dictionary of numbers. These become Green Columns in the Grid.
def calculate(coords, times, meta):
import numpy as np
z = coords[:, 2]
return {
"max_height": float(np.max(z)),
"duration": float(times[-1] - times[0]),
"metric": meta['airspeed'] * 2
}
Your function must be named filter and return True to keep the track and False to hide the track
Upload your local submission.csv (via the ๐ค Submission button) to overlay your AI
predictions on the map and calculate your local mAP score.
train.csv (for local hold-out validation) alongside test tracks.
meta dictionary so you can filter by them (e.g.,
meta['cv_fold'] == 1).