amathislab/SHOT7M2
SHOT7M2 is a synthetic, hierarchical, compositional basketball behavior dataset with 7.2 million frames that demonstrate hierarchical organization of basketball actions. Based on the animation model of Starke et al., it uses neural state machines to predict future character poses. The dataset contains 4,000 clips, each with 1,800 frames, performed by a single agent executing various basketball motions. Each clip is labeled with one of four activity types: casual play, intense play, dribbling training, or no play. It provides 26‑keypoint skeletal poses and combinations of 4 activity types, 12 actions, and 14 basic motions.
Dataset description and usage context
SHOT7M2 Dataset
Overview
- Name: SHOT7M2
- Type: Synthetic, hierarchical, compositional basketball dataset
- Frames: 7.2 million
- Content: Demonstrates hierarchical structure of basketball behaviors
- Generation Method: Based on Starke et al.'s animation model, employing neural state machines (including motion prediction and gating networks) to generate future character poses
- Data Structure: Consists of 4,000 clips, each containing 1,800 frames, with a single agent performing various basketball actions
- Activity Types: 4 (casual play, intense play, dribbling training, no play)
- Action Types: 12
- Skeleton Points: 26 key points
- Combinatorial Behaviors: 14 "Movemes"
Data Format
- File Format:
.npy - Data Splits: Training and test sets in a 32/68 ratio
- Content: 3D pose and behavior annotation data
- Test Set: Includes
benchmark_labels.npyfor benchmarking - Reading Method: Use Python scripts to load
.npyfiles
Pair the dataset with AI analysis and content workflows.
Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.