This is a sample of the Bamboo Stock brazilian cultural video dataset, provided for internal evaluation purposes ONLY. The use of this data is strictly limited by the license defined below.

Any use for training, fine-tuning, or inference of AI/ML models, or any commercial activity, is strictly prohibited with this sample.

Dataset Description

The Bamboo Stock brazilian cultural video dataset sample is a high-quality, ethically-sourced collection of 200 video clips and 2400 frames representing authentic Brazilian reality. As part of our mission to promote data sovereignty in the Global South, this dataset provides a high-fidelity, legally-sound alternative to web-scraped data.

Unlike most datasets, every asset in this collection is backed by:

IA-Ready Model Releases: Signed addendums from all actors and models specifically authorizing use for AI/ML training.
Filmmaker Agreements: Contracts ensuring all creators are compensated for the use of their work in AI models.

This dataset is specifically designed for the fine-tuning of large-scale generative video models. It provides the high-fidelity, legally-sound, and culturally rich data source required to specialize and adapt these models to new domains

Key Features

100% Human-Reviewed: All data, descriptions, and annotations are created and validated by a human-in-the-loop process, as proven by the "review" block in each JSON.
Translating Culture into Data: Our mission is to translate complex cultural realities into structured data. Our metadata goes beyond simple labels to include rich context (e.g., semantics.cultural_context).
Multimodal Data: Each video clip (.mp4) is paired with a rich descriptive JSON file.
Rich Metadata: Includes detailed scene, subject, technical, and semantic annotations.
High-Quality: Professionally shot and curated 4K cinematic footage.

Dataset Objectives & Collaboration

Our primary objective is to empower the next generation of Generative AI and Computer Vision systems with ethically sourced, high-fidelity training data. By providing a fully licensed library of authentic Brazilian reality,we aim to solve the scarcity of compliant assets in the AI landscape. We seek to connect with AI Research Labs, Foundation Model builders, and Enterprise Data Teams who prioritize legal safety, cultural nuance, and fair compensation for creators within their development pipelines.

Data Structure

The dataset is organized as follows:

/LICENSE.md - The full text of our evaluation license.
/sample_metadata.csv - A master CSV file summarizing all clips in the sample.
/taxonomy.md - A description and dictionary for all 7 categories.
/data/ - A directory containing all media files, organized by category:
- /data/Aerial/
- /data/Agriculture/
- /data/Animals/
- /data/Cities/
- /data/Food_&_Drinks/
- /data/Nature/
- /data/People/

Inside each category folder (e.g., /data/Aerial/), you will find the corresponding video and JSON pairs (e.g., 0001.mp4, 0001.json) and a /frame folder with jpeg files (e.g., 0001_001.jpeg, 0001_002.jpeg) extracted from each video, at one-second intervals

JSON Data Fields

Our metadata schema is structured as follows. (Note: null is used for fields that are not applicable to a specific clip, e.g., actors in a scene with no people).

id: Unique identifier for the clip.
source: Technical specifications of the original video file (duration, fps, resolution) and capture device (camera, lens, lighting).
metadata: Core descriptive annotations, including:
- language: Language of the annotations (e.g., "English", "pt-BR").
- country, state, city: Geolocation of the capture, if applicable.
- location_context: Description of the location type (e.g., "Indoor industrial setting").
- scene: The primary block containing the narrative description, category, and tags.
subjects: A list of human subjects (actors) present in the scene, with attributes like role, gender, age_range, and specific actions.
objects: (If applicable) A list of non-human objects with their labels, bounding boxes, and actions.
semantics: Deeper contextual understanding of the scene:
- cultural_context: Notes on symbolism, traditions, or local meaning.
- potential_use_cases: Suggested applications for the data.
review: Provenance of the annotation, confirming its status (Completed), method (Human), and timestamp.
legal: The specific license and rights information for the asset, including model release status and usage terms.

License: Bamboo Sample Evaluation License

The use of this sample dataset is governed by the following terms.

Type: Bamboo Sample Evaluation License

Scope: Non-exclusive, non-transferable license for internal evaluation only.

Permitted Uses:

Internal evaluation of the data structure, format, and quality.

Viewing for the sole purpose of assessing dataset suitability.

Testing data loading and parsing routines.

Prohibited Uses:

Any commercial use or profit-generating activity.

Training, fine-tuning, or inference of any machine learning or artificial intelligence model.

Public display, performance, distribution, or sharing.

Sublicensing, resale, or transfer to any third party.

Incorporation into any product, service, software, or application.

Creation of derivative works.

Use in any defamatory, pornographic, political, or illegal context.

Any use not explicitly listed in 'permitted_uses'.

Notice: This is a sample file for evaluation purposes only. All assets are protected by copyright. Full licensing is required for any productive use.

Citation

To license the full, production-ready dataset or to learn more about our process, please visit Bamboo Stock or contact tico@bamboostock.net.

Downloads last month: 31