seamless_interaction/src/seamless_interaction.egg-info/PKG-INFO

Metadata-Version: 2.2
Name: seamless-interaction
Version: 0.1.0
Summary: Library for seamless interaction dataset
Author: Meta AI | FAIR Seamless Team
License: Attribution-NonCommercial 4.0 International
        
        =======================================================================
        
        Creative Commons Corporation ("Creative Commons") is not a law firm and
        does not provide legal services or legal advice. Distribution of
        Creative Commons public licenses does not create a lawyer-client or
        other relationship. Creative Commons makes its licenses and related
        information available on an "as-is" basis. Creative Commons gives no
        warranties regarding its licenses, any material licensed under their
        terms and conditions, or any related information. Creative Commons
        disclaims all liability for damages resulting from their use to the
        fullest extent possible.
        
        Using Creative Commons Public Licenses
        
        Creative Commons public licenses provide a standard set of terms and
        conditions that creators and other rights holders may use to share
        original works of authorship and other material subject to copyright
        and certain other rights specified in the public license below. The
        following considerations are for informational purposes only, are not
        exhaustive, and do not form part of our licenses.
        
             Considerations for licensors: Our public licenses are
             intended for use by those authorized to give the public
             permission to use material in ways otherwise restricted by
             copyright and certain other rights. Our licenses are
             irrevocable. Licensors should read and understand the terms
             and conditions of the license they choose before applying it.
             Licensors should also secure all rights necessary before
             applying our licenses so that the public can reuse the
             material as expected. Licensors should clearly mark any
             material not subject to the license. This includes other CC-
             licensed material, or material used under an exception or
             limitation to copyright. More considerations for licensors:
            wiki.creativecommons.org/Considerations_for_licensors
        
             Considerations for the public: By using one of our public
             licenses, a licensor grants the public permission to use the
             licensed material under specified terms and conditions. If
             the licensor's permission is not necessary for any reason--for
             example, because of any applicable exception or limitation to
             copyright--then that use is not regulated by the license. Our
             licenses grant only permissions under copyright and certain
             other rights that a licensor has authority to grant. Use of
             the licensed material may still be restricted for other
             reasons, including because others have copyright or other
             rights in the material. A licensor may make special requests,
             such as asking that all changes be marked or described.
             Although not required by our licenses, you are encouraged to
             respect those requests where reasonable. More considerations
             for the public:
            wiki.creativecommons.org/Considerations_for_licensees
        
        =======================================================================
        
        Creative Commons Attribution-NonCommercial 4.0 International Public
        License
        
        By exercising the Licensed Rights (defined below), You accept and agree
        to be bound by the terms and conditions of this Creative Commons
        Attribution-NonCommercial 4.0 International Public License ("Public
        License"). To the extent this Public License may be interpreted as a
        contract, You are granted the Licensed Rights in consideration of Your
        acceptance of these terms and conditions, and the Licensor grants You
        such rights in consideration of benefits the Licensor receives from
        making the Licensed Material available under these terms and
        conditions.
        
        
        Section 1 -- Definitions.
        
          a. Adapted Material means material subject to Copyright and Similar
             Rights that is derived from or based upon the Licensed Material
             and in which the Licensed Material is translated, altered,
             arranged, transformed, or otherwise modified in a manner requiring
             permission under the Copyright and Similar Rights held by the
             Licensor. For purposes of this Public License, where the Licensed
             Material is a musical work, performance, or sound recording,
             Adapted Material is always produced where the Licensed Material is
             synched in timed relation with a moving image.
        
          b. Adapter's License means the license You apply to Your Copyright
             and Similar Rights in Your contributions to Adapted Material in
             accordance with the terms and conditions of this Public License.
        
          c. Copyright and Similar Rights means copyright and/or similar rights
             closely related to copyright including, without limitation,
             performance, broadcast, sound recording, and Sui Generis Database
             Rights, without regard to how the rights are labeled or
             categorized. For purposes of this Public License, the rights
             specified in Section 2(b)(1)-(2) are not Copyright and Similar
             Rights.
          d. Effective Technological Measures means those measures that, in the
             absence of proper authority, may not be circumvented under laws
             fulfilling obligations under Article 11 of the WIPO Copyright
             Treaty adopted on December 20, 1996, and/or similar international
             agreements.
        
          e. Exceptions and Limitations means fair use, fair dealing, and/or
             any other exception or limitation to Copyright and Similar Rights
             that applies to Your use of the Licensed Material.
        
          f. Licensed Material means the artistic or literary work, database,
             or other material to which the Licensor applied this Public
             License.
        
          g. Licensed Rights means the rights granted to You subject to the
             terms and conditions of this Public License, which are limited to
             all Copyright and Similar Rights that apply to Your use of the
             Licensed Material and that the Licensor has authority to license.
        
          h. Licensor means the individual(s) or entity(ies) granting rights
             under this Public License.
        
          i. NonCommercial means not primarily intended for or directed towards
             commercial advantage or monetary compensation. For purposes of
             this Public License, the exchange of the Licensed Material for
             other material subject to Copyright and Similar Rights by digital
             file-sharing or similar means is NonCommercial provided there is
             no payment of monetary compensation in connection with the
             exchange.
        
          j. Share means to provide material to the public by any means or
             process that requires permission under the Licensed Rights, such
             as reproduction, public display, public performance, distribution,
             dissemination, communication, or importation, and to make material
             available to the public including in ways that members of the
             public may access the material from a place and at a time
             individually chosen by them.
        
          k. Sui Generis Database Rights means rights other than copyright
             resulting from Directive 96/9/EC of the European Parliament and of
             the Council of 11 March 1996 on the legal protection of databases,
             as amended and/or succeeded, as well as other essentially
             equivalent rights anywhere in the world.
        
          l. You means the individual or entity exercising the Licensed Rights
             under this Public License. Your has a corresponding meaning.
        
        
        Section 2 -- Scope.
        
          a. License grant.
        
               1. Subject to the terms and conditions of this Public License,
                  the Licensor hereby grants You a worldwide, royalty-free,
                  non-sublicensable, non-exclusive, irrevocable license to
                  exercise the Licensed Rights in the Licensed Material to:
        
                    a. reproduce and Share the Licensed Material, in whole or
                       in part, for NonCommercial purposes only; and
        
                    b. produce, reproduce, and Share Adapted Material for
                       NonCommercial purposes only.
        
               2. Exceptions and Limitations. For the avoidance of doubt, where
                  Exceptions and Limitations apply to Your use, this Public
                  License does not apply, and You do not need to comply with
                  its terms and conditions.
        
               3. Term. The term of this Public License is specified in Section
                  6(a).
        
               4. Media and formats; technical modifications allowed. The
                  Licensor authorizes You to exercise the Licensed Rights in
                  all media and formats whether now known or hereafter created,
                  and to make technical modifications necessary to do so. The
                  Licensor waives and/or agrees not to assert any right or
                  authority to forbid You from making technical modifications
                  necessary to exercise the Licensed Rights, including
                  technical modifications necessary to circumvent Effective
                  Technological Measures. For purposes of this Public License,
                  simply making modifications authorized by this Section 2(a)
                  (4) never produces Adapted Material.
        
               5. Downstream recipients.
        
                    a. Offer from the Licensor -- Licensed Material. Every
                       recipient of the Licensed Material automatically
                       receives an offer from the Licensor to exercise the
                       Licensed Rights under the terms and conditions of this
                       Public License.
        
                    b. No downstream restrictions. You may not offer or impose
                       any additional or different terms or conditions on, or
                       apply any Effective Technological Measures to, the
                       Licensed Material if doing so restricts exercise of the
                       Licensed Rights by any recipient of the Licensed
                       Material.
        
               6. No endorsement. Nothing in this Public License constitutes or
                  may be construed as permission to assert or imply that You
                  are, or that Your use of the Licensed Material is, connected
                  with, or sponsored, endorsed, or granted official status by,
                  the Licensor or others designated to receive attribution as
                  provided in Section 3(a)(1)(A)(i).
        
          b. Other rights.
        
               1. Moral rights, such as the right of integrity, are not
                  licensed under this Public License, nor are publicity,
                  privacy, and/or other similar personality rights; however, to
                  the extent possible, the Licensor waives and/or agrees not to
                  assert any such rights held by the Licensor to the limited
                  extent necessary to allow You to exercise the Licensed
                  Rights, but not otherwise.
        
               2. Patent and trademark rights are not licensed under this
                  Public License.
        
               3. To the extent possible, the Licensor waives any right to
                  collect royalties from You for the exercise of the Licensed
                  Rights, whether directly or through a collecting society
                  under any voluntary or waivable statutory or compulsory
                  licensing scheme. In all other cases the Licensor expressly
                  reserves any right to collect such royalties, including when
                  the Licensed Material is used other than for NonCommercial
                  purposes.
        
        
        Section 3 -- License Conditions.
        
        Your exercise of the Licensed Rights is expressly made subject to the
        following conditions.
        
          a. Attribution.
        
               1. If You Share the Licensed Material (including in modified
                  form), You must:
        
                    a. retain the following if it is supplied by the Licensor
                       with the Licensed Material:
        
                         i. identification of the creator(s) of the Licensed
                            Material and any others designated to receive
                            attribution, in any reasonable manner requested by
                            the Licensor (including by pseudonym if
                            designated);
        
                        ii. a copyright notice;
        
                       iii. a notice that refers to this Public License;
        
                        iv. a notice that refers to the disclaimer of
                            warranties;
        
                         v. a URI or hyperlink to the Licensed Material to the
                            extent reasonably practicable;
        
                    b. indicate if You modified the Licensed Material and
                       retain an indication of any previous modifications; and
        
                    c. indicate the Licensed Material is licensed under this
                       Public License, and include the text of, or the URI or
                       hyperlink to, this Public License.
        
               2. You may satisfy the conditions in Section 3(a)(1) in any
                  reasonable manner based on the medium, means, and context in
                  which You Share the Licensed Material. For example, it may be
                  reasonable to satisfy the conditions by providing a URI or
                  hyperlink to a resource that includes the required
                  information.
        
               3. If requested by the Licensor, You must remove any of the
                  information required by Section 3(a)(1)(A) to the extent
                  reasonably practicable.
        
               4. If You Share Adapted Material You produce, the Adapter's
                  License You apply must not prevent recipients of the Adapted
                  Material from complying with this Public License.
        
        
        Section 4 -- Sui Generis Database Rights.
        
        Where the Licensed Rights include Sui Generis Database Rights that
        apply to Your use of the Licensed Material:
        
          a. for the avoidance of doubt, Section 2(a)(1) grants You the right
             to extract, reuse, reproduce, and Share all or a substantial
             portion of the contents of the database for NonCommercial purposes
             only;
        
          b. if You include all or a substantial portion of the database
             contents in a database in which You have Sui Generis Database
             Rights, then the database in which You have Sui Generis Database
             Rights (but not its individual contents) is Adapted Material; and
        
          c. You must comply with the conditions in Section 3(a) if You Share
             all or a substantial portion of the contents of the database.
        
        For the avoidance of doubt, this Section 4 supplements and does not
        replace Your obligations under this Public License where the Licensed
        Rights include other Copyright and Similar Rights.
        
        
        Section 5 -- Disclaimer of Warranties and Limitation of Liability.
        
          a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE
             EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS
             AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF
             ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,
             IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,
             WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
             PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,
             ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT
             KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT
             ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.
        
          b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE
             TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,
             NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,
             INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,
             COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR
             USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN
             ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR
             DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR
             IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.
        
          c. The disclaimer of warranties and limitation of liability provided
             above shall be interpreted in a manner that, to the extent
             possible, most closely approximates an absolute disclaimer and
             waiver of all liability.
        
        
        Section 6 -- Term and Termination.
        
          a. This Public License applies for the term of the Copyright and
             Similar Rights licensed here. However, if You fail to comply with
             this Public License, then Your rights under this Public License
             terminate automatically.
        
          b. Where Your right to use the Licensed Material has terminated under
             Section 6(a), it reinstates:
        
               1. automatically as of the date the violation is cured, provided
                  it is cured within 30 days of Your discovery of the
                  violation; or
        
               2. upon express reinstatement by the Licensor.
        
             For the avoidance of doubt, this Section 6(b) does not affect any
             right the Licensor may have to seek remedies for Your violations
             of this Public License.
        
          c. For the avoidance of doubt, the Licensor may also offer the
             Licensed Material under separate terms or conditions or stop
             distributing the Licensed Material at any time; however, doing so
             will not terminate this Public License.
        
          d. Sections 1, 5, 6, 7, and 8 survive termination of this Public
             License.
        
        
        Section 7 -- Other Terms and Conditions.
        
          a. The Licensor shall not be bound by any additional or different
             terms or conditions communicated by You unless expressly agreed.
        
          b. Any arrangements, understandings, or agreements regarding the
             Licensed Material not stated herein are separate from and
             independent of the terms and conditions of this Public License.
        
        
        Section 8 -- Interpretation.
        
          a. For the avoidance of doubt, this Public License does not, and
             shall not be interpreted to, reduce, limit, restrict, or impose
             conditions on any use of the Licensed Material that could lawfully
             be made without permission under this Public License.
        
          b. To the extent possible, if any provision of this Public License is
             deemed unenforceable, it shall be automatically reformed to the
             minimum extent necessary to make it enforceable. If the provision
             cannot be reformed, it shall be severed from this Public License
             without affecting the enforceability of the remaining terms and
             conditions.
        
          c. No term or condition of this Public License will be waived and no
             failure to comply consented to unless expressly agreed to by the
             Licensor.
        
          d. Nothing in this Public License constitutes or may be interpreted
             as a limitation upon, or waiver of, any privileges and immunities
             that apply to the Licensor or You, including from the legal
             processes of any jurisdiction or authority.
        
        =======================================================================
        
        Creative Commons is not a party to its public
        licenses. Notwithstanding, Creative Commons may elect to apply one of
        its public licenses to material it publishes and in those instances
        will be considered the “Licensor.” The text of the Creative Commons
        public licenses is dedicated to the public domain under the CC0 Public
        Domain Dedication. Except for the limited purpose of indicating that
        material is shared under a Creative Commons public license or as
        otherwise permitted by the Creative Commons policies published at
        creativecommons.org/policies, Creative Commons does not authorize the
        use of the trademark "Creative Commons" or any other trademark or logo
        of Creative Commons without its prior written consent including,
        without limitation, in connection with any unauthorized modifications
        to any of its public licenses or any other arrangements,
        understandings, or agreements concerning use of licensed material. For
        the avoidance of doubt, this paragraph does not form part of the
        public licenses.
        
        Creative Commons may be contacted at creativecommons.org.
        
Project-URL: Homepage, https://github.com/facebookresearch/seamless_interaction
Project-URL: Bug Tracker, https://github.com/facebookresearch/seamless_interaction
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: CC BY-NC 4.0
Classifier: Operating System :: OS Independent
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: boto3>=1.38.27
Requires-Dist: datasets>=3.6.0
Requires-Dist: fsspec>=2025.3.0
Requires-Dist: jupyterlab>=4.4.3
Requires-Dist: librosa>=0.11.0
Requires-Dist: llvmlite>=0.43.0
Requires-Dist: mypy>=1.16.1
Requires-Dist: numpy>=1.26.0
Requires-Dist: opencv-python>=4.7.0.72
Requires-Dist: pandas>=1.5.1
Requires-Dist: plotly>=6.1.2
Requires-Dist: pre-commit>=4.2.0
Requires-Dist: pylance>=0.29.0
Requires-Dist: pytest>=8.4.1
Requires-Dist: ruff>=0.12.0
Requires-Dist: s3fs>=0.4.2
Requires-Dist: streamlit>=1.46.0
Requires-Dist: webdataset>=1.0.2
Requires-Dist: wget>=3.2

<div align="center">

# Seamless Interaction Dataset

<img src="./assets/banner.gif" alt="Seamless Interaction Dataset Banner" width="800px">

**A large-scale multimodal dataset of 4,000+ hours of human interactions for AI research**


<table>
<tr>
<td align="center">
<a href="https://ai.meta.com/blog/seamless-interaction-natural-conversational-dynamics/">
🖼️ Blog
</a>
</td>
<td align="center">
<a href="https://ai.meta.com/research/seamless-interaction/">
🌐 Website
</a>
</td>
<td align="center">
<a href="https://www.aidemos.meta.com/seamless_interaction_dataset">
🎮 Demo
</a>
</td>
<td align="center">
<a href="https://huggingface.co/datasets/facebook/seamless-interaction">
🤗 HuggingFace
</a>
</td>
<td align="center">
<a href="https://ai.meta.com/research/publications/seamless-interaction-dyadic-audiovisual-motion-modeling-and-large-scale-dataset">
📄 Paper
</a>
</td>
</tr>
</table>


</div>

## Overview

Human communication involves a complex interplay of verbal and nonverbal signals, essential for conveying meaning and achieving interpersonal goals.

The **Seamless Interaction Dataset** is a large-scale collection of over 4,000 hours of face-to-face interaction footage from more than 4,000 participants in diverse contexts.
This dataset enables the development of AI technologies that understand human interactions and communication, unlocking breakthroughs in:

- 🤖 Virtual agents and embodied AI
- 🎭 Natural human-computer interaction
- 📡 Advanced telepresence experiences
- 📊 Multimodal content analysis tools
- 🎬 Animation and synthetic content generation

## 🚀 Quick Start

```bash
git clone https://github.com/facebookresearch/seamless_interaction
cd seamless_interaction
pip install -e .
streamlit run src/seamless_interaction/app/Welcome.py

# if you use uv
uv sync
uv run streamlit run src/seamless_interaction/app/Welcome.py
```

Explore the dataset with our interactive browser:

**Features:**
- 🔍 **Hierarchical Navigation**: Browse by Label → Split → Batch → Interaction
- 🎲 **Random Sampling**: Discover interactions with one-click random selection
- 📥 **Download Interface**: Download specific batches with size estimation and progress tracking
- 🎬 **Video Viewer**: Side-by-side participant videos with synchronized playback
- 📊 **Data Analysis**: Overview statistics and distribution plots
- 📁 **File Management**: Organize and preview audio, JSON, and NPZ files with expandable dropdowns

### Download Options

We provide comprehensive download methods supporting all research scales and requirements:

| **Scale** | **Size** | **Method** | **Use Case** | **Script** | **Sampling** |
|-----------|----------|------------|--------------|------------|-------------|
| 🔍 **Single Example** | ~100MB | S3 | Quick exploration, understanding data structure | [`download_s3.py`](./scripts/download_s3.py#L10) | Auto-sample from preferred vendors |
| 👥 **Interaction Pair** | ~200MB | S3 | Study conversational dynamics between participants | [`download_s3.py`](./scripts/download_s3.py#L34) | Auto-detect conversation pairs |
| 📂 **Sample Set** | ~1GB | S3/HF | Initial prototyping, algorithm development | [`download_s3.py`](./scripts/download_s3.py#L66), [`download_hf.py`](./scripts/download_hf.py#L10) | File selection or archive-based |
| 🎯 **Session Groups** | ~400MB | S3 | Deep conversational context, session dynamics | [`download_s3.py`](./scripts/download_s3.py#L100) | Auto-sample rich sessions |
| 📦 **Single Batch** | ~50GB | HF | Substantial local development, full exploration | [`download_hf.py`](./scripts/download_hf.py#L24) | WebDataset tarball download |
| 🗂️ **Multiple Batches** | ~150GB+ | HF | Training datasets, large-scale analysis | [`download_hf.py`](./scripts/download_hf.py#L38) | WebDataset tarball download |
| 🎯 **Different Splits** | Variable | HF | Cross-validation (train/dev/test, improvised/naturalistic) | [`download_hf.py`](./scripts/download_hf.py#L55) | WebDataset tarball download |
| 🌍 **Whole Dataset** | ~27TB | HF | Complete research dataset, production systems | [`download_hf.py`](./scripts/download_hf.py#L82) | WebDataset tarball download |

#### 🔍 **S3 Download** - Fine-grained exploration
Perfect for exploring individual interactions or specific file IDs. Downloads from S3 and automatically converts to consistent format (.wav, .mp4, .npz, .json).

```python
# Initialize with configuration for cleaner setup
from seamless_interaction.fs import SeamlessInteractionFS, DatasetConfig

config = DatasetConfig(
    label="improvised", 
    split="dev", 
    preferred_vendors_only=True,
    local_dir=Path.home() / "datasets/seamless_interaction",  # note: we will automatically create the directory if it doesn't exist
)
fs = SeamlessInteractionFS(config=config)
# Or use defaults: fs = SeamlessInteractionFS()

file_ids = fs.sample_random_file_ids(num_samples=1)
fs.gather_file_id_data_from_s3(file_ids[0])
# Or specify exact file: fs.gather_file_id_data_from_s3("V00_S0809_I00000582_P0947")

# Files are organized as:
# local_dir/improvised/train/0000/0005/V00_S0809_I00000582_P0947.wav
# local_dir/improvised/train/0000/0005/V00_S0809_I00000582_P0947.mp4  
# local_dir/improvised/train/0000/0005/V00_S0809_I00000582_P0947.json
# local_dir/improvised/train/0000/0005/V00_S0809_I00000582_P0947.npz
```

For more details, please refer to the [S3 Download Example](./scripts/download_s3.py).

#### 📦 **HuggingFace Download** - Batch exploration
Ideal for downloading self-contained batches (~50GB each) for local exploration. Each batch contains complete interaction pairs.

```python
from seamless_interaction.fs import SeamlessInteractionFS, DatasetConfig

# Initialize with configuration
config = DatasetConfig(label="improvised", split="dev")
fs = SeamlessInteractionFS(config=config)

# Sample set (~1GB) - Quick exploration on laptops
fs.download_batch_from_hf(batch_idx=0, archive_list=[0])

# Single batch (~50-100GB) - Substantial local development
fs.download_batch_from_hf(batch_idx=0)
```

For more details, please refer to the [HuggingFace Download Example](./scripts/download_hf.py).

### Working with Downloaded Data

```python
from seamless_interaction.fs import SeamlessInteractionFS
import json
import numpy as np
import cv2
import librosa

# Load interaction data
def load_interaction_data(file_id):
    """Load all modalities for a given file ID."""
    
    fs = SeamlessInteractionFS()
    paths = fs.get_path_list_for_file_id_local(file_id)
    print(paths)
    
    data = {}
    for path in paths:
        if path.endswith('.mp4'):
            data['video'] = cv2.VideoCapture(path)
        elif path.endswith('.wav'):
            data['audio'], data['sample_rate'] = librosa.load(path, sr=48_000)
        elif path.endswith('.json'):
            with open(path) as f:
                data['json'] = json.load(f)
        elif path.endswith('.npz'):
            data['npz'] = np.load(path)
    
    return data

fs.download_archive_from_hf(
    archive=0,
    label="improvised",
    split="test",
    batch=0,
    local_dir=None,
    extract=True,
)
# Example usage
interaction = load_interaction_data("V01_S0223_I00000127_P1505")
print(f"Available feature keys: {list(interaction['npz'].keys())}")
print(f"Right hand pose data shape: {interaction['npz']['smplh:right_hand_pose'].shape}")
```

### Basic Data Loading (HF + WebDataset)

```python
from datasets import load_dataset

# configure
label = "improvised"
split = "dev"
batch_idx = 0
archive_list = [0, 1]

base_url = (
    f"https://huggingface.co/datasets/facebook/"
    f"seamless-interaction/resolve/main/{label}/{split}/"
    "{batch_idx:04d}/{archive_idx:04d}.tar"
)
urls = [base_url.format(batch_idx=batch_idx, archive_idx=archive_idx) for archive_idx in archive_list]
dataset = load_dataset(
    "webdataset", data_files={split: urls}, split=split, streaming=True
)

for item in dataset:
    break

isinstance(item["mp4"], bytes)
# True
item["npz"].keys()
# dict_keys(['boxes_and_keypoints:box', 'boxes_and_keypoints:is_valid_box', 'boxes_and_keypoints:keypoints', 'movement:EmotionArousalToken', 'movement:EmotionValenceToken', 'movement:FAUToken', 'movement:FAUValue', 'movement:alignment_head_rotation', 'movement:alignment_translation', 'movement:emotion_arousal', 'movement:emotion_scores', 'movement:emotion_valence', 'movement:expression', 'movement:frame_latent', 'movement:gaze_encodings', 'movement:head_encodings', 'movement:hypernet_features', 'movement:is_valid', 'smplh:body_pose', 'smplh:global_orient', 'smplh:is_valid', 'smplh:left_hand_pose', 'smplh:right_hand_pose', 'smplh:translation'])
item["json"].keys()
# dict_keys(['id', 'metadata:transcript', 'metadata:vad'])
item["wav"].keys()
# dict_keys(['path', 'array', 'sampling_rate'])
```

Check out the [dataloader_webdataset.py](./scripts/dataloader_webdataset.py) script for more details.


## 🔍 Description

The `seamless_interaction` repository is split into several main components:

### 📊 Dataset

The repository provides comprehensive tools for downloading, processing, and utilizing the Seamless Interaction dataset for research and development. The dataset includes:

- **Raw and processed multimodal data**: Video, audio, transcripts, and annotations
- **Precomputed features**: Motion capture, facial keypoints, voice activity detection
- **Metadata**: Participant personality (BFI-2), interaction contexts, and relationships

### 📂 Repository Structure

```
seamless_interaction/
├── assets/               # Static assets for documentation
│   ├── banner.png
│   └── filelist.csv      # File list for the dataset
├── scripts/              # Example scripts for dataset usage
│   ├── dataloader_webdataset.py
│   ├── download_hf.py
│   ├── download_s3.py
├── src/seamless_interaction/  # Main package source code
│   ├── app/             # Data exploration application
│   ├── fs.py            # Filesystem interface for dataset access
│   ├── utils.py         # General utility functions
│   ├── constants.py     # Dataset constants and configuration
│   └── __init__.py
├── LICENSE              # CC-BY-NC 4.0 license
└── pyproject.toml       # Python package configuration
```

## 📦 Deep Dive into the Dataset

### Dataset Structure

The Seamless Interaction Dataset is organized into two main categories/labels:
- **Improvised**: Interactions primarily based on predefined scenarios with guided prompts with at least one professional actor.
- **Naturalistic**: Prompted conversations that can be carried out by normal people.

```
seamless_interaction
├── improvised                # Interactions with guided prompts
│   ├── dev
│   │   ├── 1P-IS/            # First-party internal state annotations
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.json
│   │   ├── 1P-R/             # First-party internal state rationale annotations
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.json
│   │   ├── 3P-IS/            # Third-party internal state annotations
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.json
│   │   ├── 3P-R/             # Third-party internal state rationale annotations
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.json
│   │   ├── 3P-V/             # Third-party visual annotation
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.json
│   │   ├── audio/            # Speaker-bleed denoised audio
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.wav
│   │   ├── boxes_and_keypoints/
│   │   │   ├── box/          # Bounding boxes for each participant
│   │   │   ├── is_valid_box/ # Whether bounding boxes are valid
│   │   │   └── keypoints/    # Detected facial/body keypoints
│   │   ├── movement/         # Quantified Imitator movement features
│   │   │   ├── emotion_arousal/           # Arousal measures
│   │   │   ├── emotion_valence/           # Valence measures
│   │   │   ├── emotion_scores/            # Emotion detection scores
│   │   │   ├── expression/                # Facial expression parameters
│   │   │   ├── FAUToken/                  # Facial Action Unit tokens
│   │   │   ├── FAUValue/                  # Facial Action Unit values
│   │   │   ├── gaze_encodings/            # Eye gaze direction encodings
│   │   │   ├── head_encodings/            # Head position/rotation encodings
│   │   │   ├── frame_latent/              # Per-frame latent representations
│   │   │   └── is_valid/                  # Validity flags for extracted features
│   │   ├── smplh/            # SMPL-H body model parameters
│   │   │   ├── body-pose/    # Body pose parameters
│   │   │   ├── global_orient/ # Global orientation parameters
│   │   │   ├── is_valid/     # Valid frames indicators
│   │   │   ├── left_hand_pose/ # Left hand pose parameters
│   │   │   ├── right_hand_pose/ # Right hand pose parameters
│   │   │   └── translation/  # Global translation parameters
│   │   ├── transcript/       # Time-aligned speech transcription
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.jsonl
│   │   ├── vad/              # Voice activity detection
│   │   │   └── V<vendor>_S<session>_I<interaction>_P<participant>.jsonl
│   │   └── video/            # Raw HD video recordings
│   │       └── V<vendor>_S<session>_I<interaction>_P<participant>.mp4
│   ├── test/                 # Test split with similar structure
│   └── train/                # Training split with similar structure
└── naturalistic/             # Spontaneous conversations
    ├── dev/                  # Same structure as improvised/dev
    ├── test/                 # Same structure as improvised/test
    └── train/                # Same structure as improvised/train
```

Each file is named according to a consistent convention:
- `V<vendor_id>`: Collection site/vendor identifier
- `S<session_id>`: Unique session identifier
- `I<interaction_id>`: Specific interaction within a session
- `P<participant_id>`: Individual participant identifier

### Available Modalities and Features

Each interaction in the dataset includes:

| Modality | Description | File Format | Sample Rate |
|----------|-------------|-------------|-------------|
| 🎥 Video | High-definition face-to-face footage | MP4 (H.264) | 30/29.97 FPS, 1080p |
| 🎙️ Audio | Denoised audio with separate channels | WAV | 48kHz, 16-bit |
| 📝 Transcript | Time-aligned speech transcription | JSONL | - |
| 🏃 SMPL-H | 3D body model parameters | NPY | 30 Hz |
| 🧠 Imitator Movement Features | Comprehensive quantified imitator movement data | NPY | 30 Hz |
| 📊 Annotations | Human-annotated behavioral data | JSON | - |
| 🔊 VAD | Voice activity detection | JSONL | 100 Hz |
| 📦 Keypoints | Face and body keypoints | NPY | 30 Hz |

#### Annotation Types

The dataset includes several types of human annotations for rich behavioral analysis:

| Annotation | Hours | Total Annotations | Mean # Tokens |
|------------|-------------|--------|--------|
| 1P-IS (1st-party internal state annotations) | 1.1 | 751 | 5.8 |
| 1P-R (1st-party internal state rationale annotations) | 1.1 | 751 | 10.2 |
| 3P-IS (3rd-party internal state annotations) | 4.7 | 5132 | 5.2 |
| 3P-R (3rd-party internal state rationale annotations) | 4.7 | 5132 | 11.3 |
| 3P-V (3rd-party visual annotation) | 4.7 | 5132 | 14.6 |

Please refer to the [technical report](https://ai.meta.com/research/publications/seamless-interaction-dyadic-audiovisual-motion-modeling-and-large-scale-dataset/) for a more detailed overview of annotations.

#### Movement/Imitator Feature Types

The movement directory contains rich behavioral features (output of the Imitator model):

| Feature | Description |
|---------|-------------|
| `emotion_arousal` | Arousal intensity measurements |
| `emotion_valence` | Valence (positive/negative) measurements |
| `emotion_scores` | Detected emotion categorical scores |
| `expression` | Parametric facial expression encodings |
| `FAUToken`/`FAUValue` | Facial Action Unit tokens and intensity values |
| `gaze_encodings` | Neural encodings of gaze direction |
| `head_encodings` | Neural encodings of head position and rotation |
| `frame_latent` | Per-frame latent representations |
| `alignment_head_rotation` | Head rotation data for temporal alignment |
| `alignment_translation` | Translation parameters for temporal alignment |
| `EmotionArousalToken`/`EmotionValenceToken` | Discretized emotion tokens |
| `hypernet_features` | Features from hypernetwork processing |

### Download Strategy Guide

We provide two complementary download methods optimized for different research workflows:

| Method | Use Case | Best For | Download Size | Parallelization |
|--------|----------|----------|---------------|-----------------|
| **S3 Direct** | Fine-grained exploration | Individual interactions, interaction pairs | Per file (~100MB) | ✅ Multiprocessing |
| **HuggingFace Batches** | Batch processing | Local dataset exploration, model training | ~50-100GB per batch | ✅ Multiprocessing |

#### When to Use S3 Download
- **Qualitative analysis**: Examining specific interactions in detail
- **Pair studies**: Analyzing conversational dynamics between participants
- **Feature exploration**: Understanding data structure before large downloads
- **Development**: Testing code with minimal data

#### When to Use HuggingFace Download
- **Model training**: Need substantial training data
- **Batch processing**: Analyzing patterns across many interactions
- **Local exploration**: Want self-contained dataset on laptop/workstation
- **Reproducible research**: Ensure consistent data splits

#### Performance Optimization

```python
# Optimal settings for different systems
config_default = DatasetConfig()  # auto-detects system resources
config_laptop = DatasetConfig(num_workers=4)      # Laptop/small workstation
config_workstation = DatasetConfig(num_workers=8) # High-end workstation
config_server = DatasetConfig(num_workers=16)     # Server/cluster node

# Memory-efficient batch processing
config = DatasetConfig(label="improvised", split="train")
fs = SeamlessInteractionFS(config=config)
for batch_idx in range(10):  # Process in chunks
    fs.download_batch_from_hf(batch_idx=batch_idx)
    # Process batch here...
    # Delete batch to free space if needed
```

### Dataset Versions

The dataset is organized in self-contained batches for flexible exploration:

| Split | Batches | Size per Batch | Total Size | Description |
|-------|---------|----------------|------------|-------------|
| **dev** | 5 | ~50GB | ~500GB | Development/validation set |
| **test** | 5 | ~50GB | ~500TB | Hold-out test set |
| **train** | 200+ | ~50GB | ~20TB+ | Full training data |

#### Recommended Download Strategies

```python
# Strategy 1: Quick Start (Laptop-friendly)
config = DatasetConfig(label="improvised", split="dev")
fs = SeamlessInteractionFS(config=config)
fs.download_batch_from_hf(batch_idx=0, archive_list=[0, 1, 2])  # ~6GB

# Strategy 2: Research Dataset (Workstation)
config = DatasetConfig(label="improvised", split="dev")
fs = SeamlessInteractionFS(config=config)
fs.download_batch_from_hf(batch_idx=0)     # Full dev set ~50-100GB

config = DatasetConfig(label="naturalistic", split="dev")
fs = SeamlessInteractionFS(config=config)
fs.download_batch_from_hf(batch_idx=0)   # Both interaction types

# Strategy 3: Production Training (Server/Cluster)
config = DatasetConfig(label="improvised", split="train")
fs = SeamlessInteractionFS(config=config)
for batch_idx in range(20):  # First 20 training batches (~1TB)
    fs.download_batch_from_hf(batch_idx=batch_idx)
```

#### File Format Specifications

Our data is stored in the following formats for optimal usability:

| Format | Description | Usage |
|--------|-------------|-------|
| NPZ | NumPy array files | Efficient storage of numerical feature vectors, keypoints, and parameters |
| JSONL | JSON Lines | Time-aligned annotations with one event per line (e.g., transcripts, VAD) |
| JSON | JavaScript Object Notation | Structured metadata and annotations with timestamps |
| MP4 | MPEG-4 Part 14 | High-quality compressed video with H.264 encoding |
| WAV | Waveform Audio | Uncompressed audio for highest fidelity processing |

## 🧪 Research Applications

The Seamless Interaction Dataset enables research across multiple domains:

### Embodied AI and Virtual Agents
- Train agents that display natural gestures
- Model turn-taking dynamics and interaction rhythms
- Generate contextually appropriate responses to human behavior

### Multimodal Understanding
- Analyze cross-modal correlations between speech, gesture, and expressions
- Extract behavioral patterns from large-scale interaction data
- Develop models to understand social dynamics

### Human-Computer Interaction
- Design interfaces that respond to subtle human cues
- Improve telepresence technologies with better behavioral modeling
- Create more natural conversational agents

### Animation and Content Creation
- Generate realistic human behaviors for animated characters
- Synthesize conversational dynamics for virtual production
- Create training data for digital human technologies

## ⚠️ Known Limitations and Noise in Metadata

Given the scale and complexity involved in collecting the Seamless Interaction dataset, there are several known limitations that we will address in our  ongoing work, with improvements planned for in future versions:

### Errors in Human-Based Time-Stamping
The core unit of the dataset is interactions. An interaction defines the *active time* during which a  participant’s conversation and behavior can be linked to a pair of prompts. We have observed instances of misaligned time-stamps, including:
- Annotated start/end times may be too early or too late.
- Occasional misalignment between prompt text and spoken material.
- Ordering of prompts that may contain off-by-one errors.

Despite our efforts to automatically identify and correct these errors, approximately 10% of the interactions remain affected.

### Time Stamping "Noise" in Moments of Interest (MOI)
While defining a MOI inherently involves some subjectivity, there are rare instances where:
- The described behavior only represents a subset of the observed behavior.
- The duration of the MOI does not fully capture the annotated behavior.

### Incorrect Assignment of Participant IDs
In rare instances, we have observed:
- Duplicate participant identifiers being assigned to different individuals.
- The same individual being mapped to different identifiers.

### Unreleased "Meta Time"
Currently, the dataset only contains *active time* segments - time in which two participants are actively responding to prompts. *Meta time* refers to the time between *active segments* in which participants are studying their new prompts, taking a break, etc. *Meta time* constitutes hundreds of hours in the raw collection and maybe be explored for future releases.

### Variation in Recording Site Consistency
This multi-site project contains variation in:
- Recording quality, including issues like speaker bleed and participants staying in frame.
- Acting quality in *Improvised* segments.
- The likelihood of time-stamping errors.

All vendors met our technical requirements; however,there is noticeable variation in production quality across different sites.

## 🤝 Contributing

We welcome contributions from the research community! Here are some ways to contribute:

- **Bug Reports & Feature Requests**: Open issues on GitHub
- **Dataset Improvements**: Help enhance our preprocessing pipelines or annotations
- **Model Contributions**: Submit your models to our benchmarks
- **Documentation**: Improve our guides, tutorials, and API documentation
- **Sample Code**: Share example applications built with the dataset

Please see our [CONTRIBUTING.md](CONTRIBUTING.md) for detailed guidelines, code of conduct, and submission processes.

## 📄 License & Data Usage Policy

The Seamless Interaction Dataset is licensed under CC-BY-NC 4.0 (Creative Commons Attribution-NonCommercial 4.0 International).

This means you are free to:
- **Share** — copy and redistribute the material in any medium or format
- **Adapt** — remix, transform, and build upon the material

Under the following terms:
- **Attribution** — You must give appropriate credit, provide a link to the license, and indicate if changes were made.
- **NonCommercial** — You may not use the material for commercial purposes without explicit permission.


## 📑 Citation

If you use the Seamless Interaction Dataset in your research, please cite:

<details>
<summary>BibTeX</summary>

```bibtex
@article{seamless_interaction,
  title={Seamless Interaction: Dyadic Audiovisual Motion Modeling and Large-Scale Dataset},
  author={Vasu Agrawal and
		Akinniyi Akinyemi and
		Kathryn Alvero and
		Morteza Behrooz and
		Julia Buffalini and
		Fabio Maria Carlucci and
		Joy Chen and
		Junming Chen and
		Zhang Chen and
		Shiyang Cheng and
		Praveen Chowdary and
		Joe Chuang and
		Antony D'Avirro and
		Jon Daly and
		Ning Dong and
		Mark Duppenthaler and
		Cynthia Gao and
		Jeff Girard and
		Martin Gleize and
		Sahir Gomez and
		Hongyu Gong and
		Srivathsan Govindarajan and
		Brandon Han and
		Sen He and
		Denise Hernandez and
		Yordan Hristov and
		Rongjie Huang and
		Hirofumi Inaguma and
		Somya Jain and
		Raj Janardhan and
		Qingyao Jia and
		Christopher Klaiber and
		Dejan Kovachev and
		Moneish Kumar and
		Hang Li and
		Yilei Li and
		Pavel Litvin and
		Wei Liu and
		Guangyao Ma and
		Jing Ma and
		Martin Ma and
		Xutai Ma and
		Lucas Mantovani and
		Sagar Miglani and
		Sreyas Mohan and
		Louis-Philippe Morency and
		Evonne Ng and
		Kam-Woh Ng and
		Tu Anh Nguyen and
		Amia Oberai and
		Benjamin Peloquin and
		Juan Pino and
		Jovan Popovic and
		Omid Poursaeed and
		Fabian Prada and
		Alice Rakotoarison and
		Alexander Richard and
		Christophe Ropers and
		Safiyyah Saleem and
		Vasu Sharma and
		Alex Shcherbyna and
		Jia Shen and
		Jie Shen and
		Anastasis Stathopoulos and
		Anna Sun and
		Paden Tomasello and
		Tuan Tran and
		Arina Turkatenko and
		Bo Wan and
		Chao Wang and
		Jeff Wang and
		Mary Williamson and
		Carleigh Wood and
		Tao Xiang and
		Yilin Yang and
		Zhiyuan Yao and
		Chen Zhang and
		Jiemin Zhang and
		Xinyue Zhang and
		Jason Zheng and
		Pavlo Zhyzheria and
		Jan Zikes and
		Michael Zollhoefer
  },
  url={https://ai.meta.com/research/publications/seamless-interaction-dyadic-audiovisual-motion-modeling-and-large-scale-dataset/},
  year={2025}
}
```
</details>

## 🙏 Acknowledgments

This project was made possible thanks to contributions from:

- The thousands of participants who provided interaction data
- Our dedicated annotation and QA team
- Research collaborators from multiple institutions
- FAIR (Fundamental AI Research)
- The open-source community for valuable tools and libraries
- Our data collection partners across multiple sites
- Meta Reality Labs for supporting this research initiative