In model inference tasks, when handling non-axis-aligned rotated objects (such as aircraft and ships), standard rectangular detection boxes are often not precise enough. The DeepSeek Harness (DSH) plugin ecosystem supports extending inference capabilities through plugins. xby-obb is a plugin focused on rotated object detection. It uses the YOLO26 model to output rotated bounding boxes and angles, enabling recognition that better conforms to object outlines.

Plugin Introduction

xby-obb is a DSH plugin maintained by xby-skill and belongs to the model inference tool category. It has built-in detection capabilities for 15 object categories, including plane, ship, storage tank, baseball diamond, tennis court, basketball court, ground track field, harbor, bridge, large vehicle, small vehicle, helicopter, roundabout, soccer ball field, and swimming pool. With this plugin, you can directly recognize and annotate rotated objects within DSH.

Core Features

The plugin provides the following core capabilities:

  1. Configure API Key
    Use the set_xby_apikey command to set the API key in chat. This command supports automatic persistence, so the configuration remains valid after restarting.

  2. Rotated Object Detection
    It provides three data input interfaces, all of which output rotated bounding boxes, angles, confidence scores, and class labels:
    - yolo26_obb_detect: Requires an image file URL/link.
    - yolo26_obb_detect_for_data_base64: Requires BASE64-encoded image data.
    - yolo26_obb_detect_for_data_file: Requires a local image file path.

Installation and Enablement

Add the plugin using the officially recommended GitHub installation method:

dsh plugin --profile web add github:xby_skill/xby-obb

Typical Usage

After installation, perform the following steps in a DSH session:

  1. Enter set_xby_apikey in the conversation, then enter the obtained API key to complete the configuration.
  2. Select the corresponding detection interface based on the input data type:
    - If the input is an image link, use yolo26_obb_detect.
    - If the input is BASE64 encoding, use yolo26_obb_detect_for_data_base64.
    - If the input is a local file path, use yolo26_obb_detect_for_data_file.

Use Cases and Notes

This plugin is suitable for rotated object recognition tasks that require precise bounding boxes around object outlines. Before use, note the following:
- You need to register on the official Xiaobenyang website and obtain an API key.
- The plugin follows the MIT license.

Summary

xby-obb provides DSH with the ability to directly call rotated object detection models, making it suitable for recognizing non-axis-aligned objects in remote sensing images or specific industrial scenarios. For more details and source code, refer to the plugin directory and the GitHub repository.