Checklist OPEIA 2026 Virtual Phase · CV task
Crowd Counting
Count the people in urban-scene images using point-annotated training images.
The task
Training: 100 images with head-point annotations (.mat files) and per-image counts; test: 216 images without annotations. The page suggests that detection baselines (for example YOLOv8) struggle with dense crowds and that density-estimation methods (CSRNet, CAN) perform better.
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
labeled.zip(images and .mat ground truth),test.zip,ground_truth_train.json,sample_submission.csv.- You submit
- CSV with
image_nameand count (216 rows). - Scoring
- RMSE of counts; about 30% public / 70% private split.
- Rules
- Solutions must be reproducible; the notebook used must also be submitted.
- Up to 5 submissions per day.
- Only the free version of ChatGPT may be used as an assistant (OPEIA page).
- Format
- OPEIA 2026, a single virtual phase on Kaggle (the OPEIA page gives 19 March – 5 April 2026; the public Kaggle competitions show a window of 23–24 April 2026).