Checklist AICC 2026 Round 3 · Task 1
Drawn Apart
Classify sketches into categories when only photographs and cartoons are labelled and all training sketches are unlabelled.
The task
Two explorers' data-collection robot switched its camera style mid-mission, first to saturated cartoons and then to pencil sketches. The task is to train a model that classifies sketches into their categories.
None of the sketches used for training are labelled. Labelled photographs and cartoons are provided in per-class folders, together with 9582 unlabelled sketches, 200 labelled validation sketches (for validation only) and 4624 test sketches to classify.
Contestants output the class name for every test sketch, using the class names of the photograph and cartoon folders.
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
- cartoon/<class>/ and photograph/<class>/ (labelled images), sketch/unlabeled/ (9582 sketches),
sketch_val/ (200 sketches) withval.csv,sketch_test/ (4624 sketches). - You submit
- CSV with columns filename (e.g. xxxxxx.jpg) and
class_name(exact class folder name, e.g.alarm_clock). - Scoring
- F1 score (Kaggle metric 'F1 Score'; averaging not stated). The baseline scores 0.4093.
- 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).
- Individual participation (maximum team size 1); 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.