🎥 Aria VRS Video Processing Tools

This project provides tools for downloading and converting Aria VRS files.

Web: https://www.projectaria.com/datasets/aea/

Doc: https://facebookresearch.github.io/projectaria_tools/

📁 File Description

• download_vrs.py: VRS file download tool
• create_video.py: VRS to MP4 video conversion tool including Debayering , Devignetting, Undistortion, Rotation, Encode.
• AriaEverydayActivities_download_urls.json: JSON file containing download links. Including not only vrs, but also other data like slam results, gazes...
• requirements.txt: Python dependencies list

🚀 Environment Setup

1. Create Python Virtual Environment and Install Dependencies

# Create virtual environment
python3 -m venv env

# Activate virtual environment
source env/bin/activate

# Install required packages
pip install -r requirements.txt

Use ffmpeg for video re-encoding (recommended for better compression and preview):

sudo apt install ffmpeg

---

📖 Usage Guide

2. 📥 Download VRS Files

Use download_vrs.py to download .vrs files from a JSON file.

Basic Usage

# Download all .vrs files using default JSON file (AriaEverydayActivities_download_urls.json)
python download_vrs.py --outdir ./aria_downloads

Advanced Options

--json_path AriaEverydayActivities_download_urls.json # if you have custom json
--workers 4 # num of doewnload workers
--max-files 10 # download partial dataset, -1 = download all

---

3. 🎬 Convert VRS Files to Video

Use create_video.py to convert a single .vrs file to MP4 video.

Basic Usage

# Generate video from camera-rgb
python create_video.py --vrs xxx.vrs --output output.mp4 --camera camera-rgb

📹 Select Different Cameras

# RGB camera (color, automatically rotated 90 degrees)
--camera camera-rgb
# SLAM left camera (grayscale)
--camera camera-slam-left
# SLAM right camera (grayscale)
--camera camera-slam-right

⚙️ Video Parameter Adjustments

# Set output video FPS
--fps 60

# Process only first 300 frames (for testing)
--max_frames 300

# Adjust ffmpeg encoding quality (CRF value: 0-51, lower is better quality, default 23)
--crf 18

# Adjust encoding speed (preset: ultrafast, fast, medium, slow, veryslow)
# Slower presets produce smaller files but take longer to encode
--preset slow

🌟 Devignetting Settings

# Change devignetting mask folder
--devignetting_mask ./dv_mask_bin

🔧 Image Correction Parameters

# Adjust undistorted image resolution and field of view
--undist_w 1024 
--undist_h 1024 
--undist_fov 150

🔄 Processing Pipeline

create_video.py processing pipeline includes the following steps:

1. 📂 Read VRS File: Open .vrs file using projectaria_tools
2. 🎨 Debayering: Convert raw Bayer pattern to RGB (RGB camera only, handled by data provider)
3. ✨ Devignetting: Remove vignetting effect if mask folder is provided (applied by data provider)
4. 🔍 Image Undistortion: Undistort each frame, converting fisheye to pinhole projection
5. 🔄 Image Rotation: Rotate RGB camera images 90 degrees clockwise (RGB camera only)
6. 💾 Write Temp Video: Write temp file using OpenCV FFV1 or MJPG codec
7. 🎞️ Re-encode: Re-encode with ffmpeg using H.264 (libx264) to significantly reduce file size

Note: SLAM cameras (grayscale) skip step 2 and 5.