Understand the project
Practical overview
When it fits
Use it when you need a starting point for training a hand detector, especially for table-top or first-person camera views, and can work with the repository’s older TensorFlow setup or adapt its model-export process.
Who it is for
- Computer-vision developers adapting hand detection to their own camera views.
- Researchers exploring neural-network hand detection with egocentric imagery.
Documented uses
- Prepare annotated hand imagery for an object-detection task.
- Train an SSD hand detector with TensorFlow’s Object Detection API.
- Apply a trained detector to webcam or video footage.
Before you choose
Requirements & limits
The repository is public and MIT-licensed.
- The project’s code was written and tested with TensorFlow 1.4.0-rc0; the README warns that other versions may produce errors.
- A suitable annotated dataset is central to training. The author found Egohands a better fit for the project’s egocentric needs than the Oxford Hands Dataset.
- The README reports 11–21 frames per second in particular MacBook Pro CPU examples, depending on image size and visualization.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Source-based practical guide
Prepare a frozen TensorFlow hand-detector graph
Export a model checkpoint as a frozen inference graph for hand detection.
Before you start
- TensorFlow Object Detection API export_inference_graph.py script available in the current directory.
- Model configuration and checkpoint files, such as the referenced SSD MobileNet files.
- The repository notes its code was written and tested with TensorFlow 1.4.0-rc0; other versions may produce errors.
Steps
Use the Object Detection API export_inference_graph.py script.
Set the input type to image_tensor.
Provide the model configuration and checkpoint paths.
Set the output directory to hand_inference_graph.
Expected outcome
A frozen graph is written to the hand_inference_graph output directory.
Guide limitations
- The documented export example may require generating a frozen graph for the TensorFlow version in use.
- Using a TensorFlow version other than 1.4.0-rc0 may result in errors.
Evidence and transparency
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- victordibia/handtracking · Official homepage title · retained in its source language