📦 Installation via pip
This guide describes how to install MSight using Python packages. This method is ideal for compact edge boxes and rapid prototyping.
🧱 Core Installation (Minimum Setup)
To get started with MSight, install the foundational packages:
pip install msight-base msight-core
This provides the minimum required setup for MSight:
-
🧱 msight-base
-
Defines fundamental data structures and abstractions
- Includes road users, points, trajectories, frames
-
Provides core utilities such as trajectory management and visualization
-
⚙️ msight-core
-
Provides the core roadside infrastructure
- Sensor data receiving and streaming
- Pub/Sub communication system
- Data serialization and pipeline orchestration
- Cloud integration capabilities
- V2X communication support
👉 This setup is sufficient for running basic MSight pipelines and infrastructure.
🎥 2D Vision Support (Camera-based Perception)
If your deployment requires camera-based perception (RGB, fisheye, infrared), install:
pip install msight-base msight-core msight-vision
This adds:
-
🎥 msight-vision
-
Image-based object detection
- Localization
- Multi-object tracking
- State estimation
👉 Required for all 2D sensor processing pipelines.
🛰️ LiDAR Support (3D Perception)
If your deployment includes LiDAR sensors, install:
pip install msight-base msight-core msight-lidar
This adds:
-
🛰️ msight-lidar
-
3D perception
- Point cloud processing
- BEV-based understanding (planned)
⚠️ Note:
msight-lidaris currently under development and may not yet be publicly available.
🧠 Installation Philosophy
MSight follows a modular installation strategy:
- Install only what you need
- Keep edge deployments lightweight
- Enable flexible combinations of sensing modalities
| Use Case | Installation |
|---|---|
| Core infrastructure | msight-base + msight-core |
| Camera perception | + msight-vision |
| LiDAR perception | + msight-lidar |
✅ Verification
After installation, verify your setup:
import msight_base
import msight_core
Optional modules:
import msight_vision
import msight_lidar