任务库

从真实的标注任务开始。

从任务场景出发，快速掌握模型配置与数据标注流程。

26个持续维护的示例查看源码

分类

Image-level Classification

Image classification assigns labels to an entire image. X-AnyLabeling supports multiclass and multilabel classification.

打开任务指南

分类

Shape-level Classification

Multi-task Classification involves training a model to perform multiple classification tasks simultaneously. For example, a model could be trained to classify both the type of person and vehicle attributes in a single image.

打开任务指南

计数

GeCo

GeCo is a low-shot counting model that uses exemplar boxes to detect, segment, and count objects.

打开任务指南

计数

GECO2

GECO2 is a few-shot object counting model designed for dense and scale-varying scenes. Instead of relying on heuristic image tiling or aggressive upscaling, it builds exemplar-aware query features across multiple backbone resolutions and fuses them into a high-resolution detection map. This makes it better suited for crowded images with small objects, large objects, or both at the same time. In the X-AnyLabeling workflow, GECO2 is exposed through the remote inference service provided by X-AnyLabeling-Server.

打开任务指南

描述

Image Captioning

X-AnyLabeling supports image-captioning workflows through its Chatbot and visual question answering tools. Both tools can send the current image to a vision-language model and store or export the generated description for review.

打开任务指南

描述

Tagging Annotation

Tagging assigns descriptive labels to an entire image or to selected shapes.

打开任务指南

检测

Object Detection

Object detection identifies objects and their locations in an image.

打开任务指南

检测

Oriented Object Detection

Oriented object detection surpasses standard object detection by adding angular precision to pinpoint objects in images.

打开任务指南

估计

Depth Estimation

Depth-estimation models estimate the relative distance from the camera for every pixel in an image.

打开任务指南

估计

Face Estimation

Face estimation covers face detection and facial landmark annotation. X-AnyLabeling supports models that create a face rectangle and five facial keypoints with a shared groupid, making it easy to keep each face and its landmarks linked as one object.

打开任务指南

估计

Pose Estimation

Pose estimation identifies keypoints such as joints, landmarks, or other distinctive object features.

打开任务指南

开放词汇定位

LocateAnything

LocateAnything is a vision-language grounding model that returns boxes and points using normalized coordinates. X-AnyLabeling integrates it through X-AnyLabeling-Server, so inference runs in a dedicated server environment and the client receives standard annotation shapes.

打开任务指南

开放词汇定位

SAM 3

打开任务指南

开放词汇定位

YOLOE

YOLOE (You Only Look Once, Extended) brings state-of-the-art "anything" detection capabilities to X-AnyLabeling. Based on the research paper YOLOE, this model enables real-time detection and segmentation of any object you can describe or indicate visually, eliminating traditional category constraints.

打开任务指南

交互式视频目标分割

SAM 2 Video Segmentation

Interactive Video Object Segmentation (iVOS) has become an essential task for efficiently obtaining object segmentations in videos, often guided by user inputs like scribbles, clicks, or bounding boxes. In this tutorial, you'll learn how to leverage the video tracking feature of SAM2 on X-AnyLabeling to accomplish iVOS tasks.

打开任务指南

交互式视频目标分割

SAM 3 Video Segmentation

SAM3 is a unified foundation model for promptable segmentation in images and videos. It can detect, segment, and track objects using text or visual prompts such as points, boxes, and masks. Compared to its predecessor SAM2, SAM3 introduces the ability to exhaustively segment all instances of an open-vocabulary concept specified by a short text phrase. SAM3 Video performs Promptable Concept Segmentation (PCS) on videos, taking text as prompts and automatically detecting and tracking all matching object instances across video frames.

打开任务指南

图像抠图

Image Matting

Image Matting is the process of accurately estimating the foreground object in images and videos. It is a very important technique in image and video editing applications, particularly in film production for creating visual effects.

打开任务指南

多目标跟踪

Multi-Object Tracking

Multi-Object Tracking (MOT) technology is used to simultaneously identify and track multiple targets within video sequences, involving the association of targets across different frames.

打开任务指南

光学字符识别

Document Layout Analysis

Document Layout Analysis is used to identify and extract the layout structure of documents, including text blocks, images, tables, and other elements.

打开任务指南

光学字符识别

Key Information Extraction

Key Information Extraction (KIE) involves extracting essential information from images, such as retrieving a name and ID number from an ID card. The type of information extracted is often specific to the task at hand but can vary across different tasks.

打开任务指南

光学字符识别

PaddleOCR-VL-1.5

PaddleOCR-VL-1.5 is a unified Vision-Language OCR model that supports multiple document understanding tasks through a single model architecture. Built upon powerful vision-language foundations, it can handle diverse OCR scenarios including text recognition, table extraction, formula recognition, chart understanding, seal recognition, and text spotting with bounding boxes.

打开任务指南

光学字符识别

Text Recognition

Currently, X-AnyLabeling supports both manual and automatic annotation of the PP-OCR dataset.

打开任务指南

分割

Image Segmentation

For example, to segment a plant while excluding its pot, place positive points on the plant and negative points on the pot. Press F when the mask is ready, enter a label, and save the object.

打开任务指南

训练

Model Training

This guide walks you through training Ultralytics models directly within the X-AnyLabeling GUI. You can prepare your dataset, configure training parameters, launch a training job, and monitor its progress—all from one convenient interface.

打开任务指南

视觉语言

Florence 2

Florence-2 is a novel vision foundation model with a unified, prompt-based representation for a variety of computer vision and vision-language tasks, developed by Microsoft.

打开任务指南

视觉语言

Rex-Omni

Rex-Omni is a 3B-parameter Multimodal Large Language Model (MLLM) that redefines object detection and a wide range of other visual perception tasks as a simple next-token prediction problem.

打开任务指南