An automated tool for estimating camera motion using optical flow and geometric transformations to classify video movements like Pan, Tilt, and Dolly.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install dynamic-object-aware-egomotion-egomotion-estimation
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install dynamic-object-aware-egomotion-egomotion-estimation using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Egomotion Estimation Skill?
The Egomotion Estimation skill is a technical implementation within the Openclaw Skills ecosystem designed to analyze and classify camera motion within video streams. By leveraging sophisticated computer vision techniques such as optical flow and affine or homography transformations, it accurately identifies complex camera movements including Stay, Dolly, Pan, Tilt, and Roll. This skill is built for developers who need robust, frame-by-frame analysis that supports multi-label assignment, ensuring that simultaneous motions like zooming while panning are captured accurately.
Egomotion Estimation Skill Use Cases
- Classifying cinematic camera movements for automated video indexing and archival.
- Generating metadata for large-scale video datasets used in machine learning training.
- Enhancing AI-driven video editing workflows by identifying specific motion triggers or patterns.
- Analyzing drone or robotics footage to understand relative camera displacement and orientation.
How Egomotion Estimation Skill Works
- Feature tracking: The skill utilizes goodFeaturesToTrack and calcOpticalFlowPyrLK to identify and track key points across frames.
- Robust transform: It applies estimateAffinePartial2D or homography using RANSAC to derive translation, rotation, and scale data.
- Thresholding: The derived values are compared against configurable thresholds for translation, rotation, and scale to assign labels.
- Temporal smoothing: A windowed median filter is applied to the raw labels to reduce flicker and high-frequency noise.
- Interval compression: Consecutive frames sharing identical label sets are merged into start-to-end intervals for efficient data representation.
Egomotion Estimation Skill Setup
To deploy this skill within your Openclaw Skills environment, ensure you have the required vision processing dependencies installed.
# Install required computer vision libraries
pip install opencv-python numpy
# Configure the skill thresholds based on your video resolution
# Default heuristics are provided for 720p footage
Egomotion Estimation Skill Data Schema & Taxonomy
The skill produces a structured data output representing movement intervals across the video timeline.
| Property | Type | Description |
|---|---|---|
| start_frame | Integer | The beginning index of the motion segment |
| end_frame | Integer | The ending index of the motion segment |
| labels | Array | A list of strings representing the detected motions (e.g., Dolly In, Pan Left) |
| metadata | Object | Contains the raw transformation matrices and RANSAC inlier ratios |
Egomotion Estimation Skill Advanced Features
- Multi-label frame support allowing for complex motion detection such as simultaneous Roll and Dolly.
- Automated interval compression to minimize data size for long-form video analysis.
- Resolution-independent normalization for translation and scale thresholds.
- RANSAC-based outlier rejection to maintain accuracy in low-texture or high-noise environments.
SKILL.md
Loading
Related Openclaw Skills
Sampling and Indexing
A standardized protocol for video sampling and frame indexing to ensure consistency across downstream AI instructions and artifacts.
Output Validation Skill
A specialized tool for local self-checking of instruction formats and mask output consistency without requiring ground truth data.
D3.js Data Visualization
A specialized skill for building deterministic, offline-ready data visualizations using D3.js from local CSV, TSV, and JSON datasets.
MoltMail: Web3 Inbox & Email Infrastructure
MoltMail provides AI agents with a secure Web3 email identity and integrated crypto wallet for autonomous communication and financial transactions.
Dialogue Graph
A comprehensive library for building, validating, and visualizing structured dialogue trees and branching narrative flows.
Wisgate Model Management
A specialized utility for querying, configuring, and updating Wisgate model metadata within the OpenClaw ecosystem.
Featured*
