Checklist AICC 2026 Round 3 · Task 2
Sound of Nature
Classify 5-second field recordings into nine animal and nature sound classes.
The task
Automated classification of natural sounds supports scalable, non-invasive environmental monitoring. Models must cope with varying recording conditions, overlapping sources and subtle acoustic differences.
The dataset contains 5-second mono WAV clips at 44.1 kHz recorded in diverse field conditions. The training set has 800 labelled clips; the test set has 200 clips, split 100/100 into public and private leaderboard parts. The nine classes are cat, chirping_birds, cow, dog, hen, pig, rain, sheep and water_drops.
Contestants predict the class name for every test clip.
Abridged by SOTA from the official materials. The official statement has the exact rules, and it wins wherever this summary differs.
At a glance
- You get
- audio/ (
{sampleID}.wav),train.csv(800 rows: sampleID, class) andtest.csv(200 rows: sampleID). - You submit
submission.csvwith columns sampleID and class; every test ID exactly once, order irrelevant.- Scoring
- Macro-averaged F1, mapped to points: below 0.3 gives 10 points; 0.3 <= F1 < 0.9 is normalised linearly between 10 and 90 points; 0.9 or above gives 100 points (as stated). Malformed submissions receive 0 points and do not count towards the submission limit.
- Rules
- No pretrained models other than those provided in torchvision.
- No training on validation or test samples in any way (including unsupervised learning or computing statistics over them).
- No external datasets; no labelling of the unlabelled sketches or the test set with a pretrained model (for example CLIP).
- Maximum team size 10 on Kaggle; at most 50 submissions per day.
- AICC contest rules (stated on each Kaggle rules page, not enforceable): no use of LLMs for writing code or getting task ideas; no internet use other than official library documentation and the contest platform; no communication with anyone during the contest; clarifications only via the #clarification-requests channel on the AICC Discord server.
- Format
- AICC Round 3, online on Kaggle, 29 Jan 2026 18:00 UTC – 31 Jan 2026 18:00 UTC.