gauduc/ulatroi

🧠 Project SLOB: Spontaneous Lifestyle & Observational Behaviors Dataset

📌 Abstract

Welcome to the primary data ingestion node for Project SLOB. This repository hosts a massive, high-fidelity multimodal dataset designed to train next-generation Artificial Intelligence in recognizing, analyzing, and predicting spontaneous human behaviors in unconstrained, real-world video streams.

This node is strictly used for the Spatial-Temporal Audio-Visual Synchronization (ST-AVS) phase of the SLOB architecture.

📂 Dataset Architecture & Modalities

This repository operates as a dynamic, rolling pipeline. You will encounter various file types which are outputs of our multi-pass processing nodes:

• Raw Video Streams (video.mp4): High-framerate visual data used for spatial tracking and behavioral bounding-box generation.
• Synthetic Audio Injections (audio_vi_tts.mp3): AI-generated audio arrays used to test the model's robustness against audio-visual desynchronization and cross-lingual hallucination.
• Time-Warped Subtitles (VI_STRETCHED.srt): Transcripts that have been deliberately dilated (e.g., 0.85x speed factor) to train the AI's temporal alignment algorithms.
• Rendered Composites (FINAL_*.mp4): The final evaluation tensors, combining scaled visual inputs, lowered-dB original audio, and synthetic overlays.

⏳ Data Retention Policy (Rolling Window)

Due to the sheer volume of data (scaling up to 8.7TB) and our strict adherence to Continuous Learning (CL) memory constraints, this repository operates on a 7-day rolling purge cycle. Data is autonomously ingested, processed by our GPU clusters, and permanently deleted after 168 hours to simulate limited-memory learning environments and comply with automated data hygiene protocols.

⚠️ Access Restrictions & Warnings

1. Automated Environment: This repository is managed 100% by automated CI/CD pipelines and headless GPU nodes.
2. Unstable Files: Files may appear corrupted or incomplete if accessed during an active tensor-writing cycle. Manual downloading is highly discouraged.
3. No Manual Pull Requests: Any unauthorized modifications, pull requests, or manual file uploads will be automatically rejected and purged by the master node.

⚖️ Ethical Considerations & Data Privacy

All audio-visual streams ingested into the SLOB framework are strictly purposed for non-commercial, academic algorithm training (Fair Use). No personally identifiable information (PII) is permanently retained. The synthetic overlays and audio augmentations are completely decoupled from the original subjects. By nature of our 7-day rolling purge mechanism, we ensure strict adherence to temporary caching policies.

📖 Citation

If you utilize the SLOB methodology in your ST-AVS research, please use the following BibTeX entry (dataset weights access required):

@dataset{slob_dataset_2026,
  author       = {SLOB Research Core},
  title        = {Spontaneous Lifestyle & Observational Behaviors (SLOB) Dataset},
  year         = {2026},
  publisher    = {Hugging Face},
  version      = {1.4.2},
  howpublished = {\url{[https://huggingface.co/datasets/gauduc/video_public](https://huggingface.co/datasets/gauduc/video_public)}}
}