---
license: mit
task_categories:
- image-classification
- tabular-regression
- tabular-classification
- reinforcement-learning
- robotics
tags:
- bci
- brain-computer-interface
- neuroscience
- gaming
- fps
- RLHF
- signal-processing
- motor-imagery
- A11Y
- WCAG
---
<div style="
    background: #00FF00;
    border-left: 4px solid #00FF00;
    padding: 1.5rem;
    margin: 2rem 0;
    font-family: 'Fira Code', 'Courier New', monospace;
    color: #00FF00;
    border-radius: 0 8px 8px 0;
">
    <pre style="
        font-size: 8px;
        line-height: 1.2;
        margin: 0;
        overflow-x: auto;
        color: #00FF00;
    ">
________      _____________________    ______________________________.___ ____  __.___________
\______ \    /  _  \__    ___/  _  \  /   _____/\__    ___/\______   \   |    |/ _|\_   _____/
 |    |  \  /  /_\  \|    | /  /_\  \ \_____  \   |    |    |       _/   |      <   |    __)_ 
 |    `   \/    |    \    |/    |    \/        \  |    |    |    |   \   |    |  \  |        \
/_______  /\____|__  /____|\____|__  /_______  /  |____|    |____|_  /___|____|__ \/_______  /
        \/         \/              \/        \/                    \/            \/        \/ 

</div>
       
# DATASTRIKE BCI Timelapse Dataset

*UNDER DEVELOPMENT*

*This dataset was created with the DATASTRIKE app located in the /gym/ folder.*

## Description

Time-synchronized multimodal dataset for BCI intent recognition, collected with frame-by-frame timelapse capture during FPS gameplay.

A time-synchronized multimodal dataset hosted on Hugging Face for brain-computer interface (BCI) intent 
ecognition research. It was collected via frame-by-frame timelapse capture during first-person shooter (FPS) gameplay and includes 
synchronized image sequences (320x240 JPGs), game state data (like player position, velocity, ammo, and combat stats), BCI intent labels 
across 13 categories, input data (mouse/keyboard), and RLHF ratings for combat and capture actions. The dataset is small, with 188 total 
frames grouped into 1 temporal sequence from a 1:35-minute session, and a download size of about 685 kB. It's structured with JSONL files 
for metadata and intents, a directory of images, and a sequences.json file for time-series analysis, making it suitable for deep learning 
models like LSTMs or Transformers on multimodal temporal data. Tags include BCI, timelapse, FPS gameplay, intent recognition, multimodal, 
time-series, RLHF, and sequence modeling.

## Key Features
- **Frame-by-Frame Timelapse**: Synchronized image sequences at 320x240 resolution
- **Multiple Capture Modes**: Manual (LMB), Auto-interval, and Sequence recording
- **BCI Intent Labels**: 13 intent categories including timelapse capture events
- **Time-Synced Game State**: Every frame includes synchronized game state data
- **RLHF Data**: Automated ratings for combat events and capture actions
- **Sequence Analysis**: Grouped frames into temporal sequences for time-series analysis

## Dataset Structure
- `timelapse_frames.jsonl`: Frame metadata (one per line)
- `frames/`: JPG images for each frame
- `sequences.json`: Temporal grouping of frames
- `bci_intents.jsonl`: BCI intent transition history
- `metadata.json`: Dataset statistics and configuration
- `README.md`: This documentation

## Capture Modes
1. **MANUAL**: LMB click captures single frame (hold for burst)
2. **AUTO**: Automatic capture at 500ms intervals
3. **SEQUENCE**: Start/stop recording for continuous frame sequences

## Dataset Statistics
- **Total Frames**: 188
- **Sequences**: 1
- **Session Duration**: 01:35
- **Player Level**: 1
- **Accuracy**: 34%
- **Total Kills**: 30

## Frame Data Structure
Each frame includes:
- Image data (320x240 JPG)
- Timestamp and game time
- BCI intent label
- Full game state (position, rotation, velocity, ammo, etc.)
- Input data (mouse movements, keyboard state)
- RLHF rating (if applicable)

## Usage for BCI Research
```python
import json
import cv2
import numpy as np

# Load frame metadata
frames = []
with open('timelapse_frames.jsonl', 'r') as f:
    for line in f:
        frames.append(json.loads(line))

# Create time-series dataset
X_images = []
X_game_state = []
y_intents = []

for frame in frames:
    # Load image
    img_path = f"frames/{frame['image_filename'].split('/')[1]}"
    img = cv2.imread(img_path)
    X_images.append(img)
    
    # Game state features
    game_state = frame['game_state']
    features = [
        game_state['player_position'][0],  # x
        game_state['player_position'][2],  # z
        game_state['combat_state']['ammo'],
        game_state['game_stats']['level']
    ]
    X_game_state.append(features)
    
    # BCI intent label
    y_intents.append(frame['bci_intent'])

# For sequence modeling
sequences = json.load(open('sequences.json', 'r'))
for seq_id, sequence in sequences.items():
    seq_frames = [f for f in frames if f['sequence_id'] == int(seq_id)]
    # Process as temporal sequence for LSTM/Transformer models
```

## Citation
```
@dataset{datastrike_bci,
  title={DATASTRIKE BCI Timelapse Dataset},
  author={webXOS},
  year={2026},
  url={https://github.com/webxos/datastrike-bci},
  note={Frame-by-frame timelapse capture for BCI intent recognition}
}
```

Generated on 2026-01-07 by DATASTRIKE BCI Timelapse v4.0 by webXOS