# Beginner's Hypothesis 2025 — Data

来源：https://www.kaggle.com/competitions/beginners-hypothesis-25/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

The dataset comprises video files and a corresponding CSV containing Titan attributes.
In Training_Data folder , there is the CSV file containing true values for each corresponding video in Train_Videos.Your task is to create a CSV file of same format for all the videos in Testing_Data.
Here's what you need to know:
**Videos:**
Each video captures the activity of a single Titan in a 3x64×64 resolution across 20 frames. 
**Attributes (Columns in the CSV):**
- **video_id:** The identifier of the video, matching the filename (e.g., 1.mp4).
- **element:** Encoded attribute that might describe the Titan's general form.
- **motion:** The Titan's movement pattern (e.g., zigzag, random).
- **power**: Encoded attribute representing a prominent feature of the Titan's appearance.
- **speed**: The speed of the Titan's movement.
- **video_summary**: A 2D representation of the whole video in some ancient cryptic way.

**Noise in the Data:**
Videos and textual attributes have partially been corrupted by addition of some noise due to intense heat emitted by the titans. You have to carefully preprocess the data to extract data from noise.