Discord

Checklist ROAI 2025 Selection Camp – Test 2 · Task 2

Skeletons Don't Lie: Can AI Decode Your Moves?

Classify the action and the camera viewpoint of skeleton sequences given as per-frame 3D joint coordinates.

  • Time series
  • Vision
  • Sequence classification (skeleton-based action recognition)
Read the official statement

The task

Each sample is a sequence of frames, and each frame gives the 3D positions of 25 body joints (J1X … J25Z) with a sequence identifier IDSample and two labels: Action (0 wear jacket, 1 take off a hat/cap, 2 hopping, 3 jump up, 4 falling) and Camera (1 45° view, 2 front view, 3 side view).

Subtask 1 asks for the number of frames of each test sequence, subtask 2 for its action and subtask 3 for its camera.

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
train_data.csv and test_data.csv with FrameNumber, IDSample and joint coordinates (labels in training only); a PDF version of the statement with the joint layout figure is in the starter kit.
You submit
output.csv with subtaskID (1–3), datapointID (IDSample) and answer.
Scoring
Subtask 1: 10 points. Subtask 2 (40 points): accuracy, full marks at ≥ 0.965. Subtask 3 (50 points): accuracy, full marks at ≥ 0.8.
Rules
  • On-site during the national training and selection camp (17–24 June 2025, POLITEHNICA Bucharest)
  • Internet limited to the contest platform and the documentation of the available libraries
  • Only submissions ticked during the round count; anti-fraud session locking on the judge
Format
2025 national selection camp, test 2, 23 June 2025.

Statement

Competition

Solution

Details

Year
2025, Bucharest, Romania (POLITEHNICA Bucharest)
Round
Selection Camp – Test 2 · Task 2
Language
English
License
Not stated by the source