Skip to content

README.md

Latest commit

 

History

History
32 lines (20 loc) · 1.54 KB

README.md

File metadata and controls

32 lines (20 loc) · 1.54 KB

GeCo Example

Introduction

GeCo (NeurIPS'24) is a novel low-shot counter that uses a unified architecture for accurate object detection, segmentation, and count estimation, robustly generalizing object prototypes and employing a novel counting loss for direct optimization of the detection task, significantly outperforming existing methods.

Usage 🚀

  1. 📁 Load your media into X-AnyLabeling:

    • 🖼️ Images: Press Ctrl+I for a single image or Ctrl+U for a folder
    • 🎥 Videos: Press Ctrl+O

  2. 🤖 Download GeCo model files manually

    Currently, due to GitHub release size limitations (>2GB), you need to manually download the GeCo model files from one of the following sources:

    After downloading, place the files in the expected directory described in the custom model documentation. [English | 中文版].

  3. ✏️ Drawing and labeling:

    • 📦 Click the Rect tool to start
    • 🎯 Draw one or more bounding boxes around objects of interest
    • ✅ Press F or click Finish when done
    • 🏷️ Enter the class name for the labeled object
    • ↩️ Made a mistake? Press B or click Clear to undo