Select a Model
=============================

.. toctree::
    :maxdepth: 1
    :hidden:

    semantic/index
    ano/index
    position/index
    presence/index
    seg/index
    kp/index
    class/index
    obj/index
    rot_obj/index
    mix_model/index
    ocr/index
    playground/index


DaoAI World supports the following models:

    - :ref:`Supervised Defect Segmentation`
    - :ref:`Unsupervised Defect Segmentation`
    - :ref:`Positioning`
    - :ref:`Presence Detection`
    - :ref:`Instance Segmentation`
    - :ref:`Keypoint Detection`
    - :ref:`Image Classification`
    - :ref:`Object Detection`
    - :ref:`Rotated Object Detection`
    - :ref:`Mixed Model`
    - :ref:`OCR`

**How to Choose a Model?** Different models have their own unique features and best-use cases. Here are some points to help you select the right model:

- Detecting objects in an image:

    - :ref:`Object Detection` : Detects the **type** and **bounding box** of an object. Use this model when you need to detect the presence of objects in an image. Example scenarios: detecting the number of vehicles on the road, checking if screws are installed or missing in a part, detecting whether workers are wearing helmets.
        
        .. image:: images/obj.png
            :scale: 100%

    - :ref:`Instance Segmentation` : Detects the **type** and **mask** of an object. Use this model when you need to detect the presence of objects and their masks. Example scenarios: identifying the position and area of a package, locating parts inside a box, or classifying different types of screws in one image.
        
        .. image:: images/insseg.png
            :scale: 100%
            
    - :ref:`Keypoint Detection` : Detects the **type** , **mask** , and **keypoints/pose** of an object. Use this model when you need instance segmentation with additional keypoint detection. Example scenarios: detecting the posture of athletes, determining the current position and rotation of an object.
        
        .. image:: images/kp.png
            :scale: 100%
            
- Image Classification:    
    - :ref:`Image Classification` : Detects the **category** of an image. Use this model when you need to classify images. Example scenarios: classifying product models, checking if specific screw holes are installed correctly, diagnosing diseases from X-rays.
        
        .. image:: images/class.png
            :scale: 100%

    - :ref:`Rotated Object Detection` : Detects the **type** and **rotated bounding box** of objects. Use this model when objects in your image are often rotated or not aligned with the axes. Example scenarios: detecting electronic components at arbitrary angles, rotated packaging on a conveyor belt, identifying tilted labels.

        .. image:: images/rot_obj.png
            :scale: 100%

    - :ref:`Mixed Model` : Performs both detection and classification of multiple object types in a single image. Use this model when you need to locate and classify various types of objects simultaneously. Example scenarios: intelligent surveillance to detect pedestrians, safety gear detection in smart factories, inspecting the quality of sunflower seeds, rice, or solder joints on chips.

        .. image:: images/mix_model.png
            :scale: 100%

- Detecting anomalies or surface defects on objects:
    
    - :ref:`Unsupervised Defect Segmentation` : Detects the **mask** of abnormal areas in an image. Use this model when you need to detect abnormal regions in an image. Example scenarios: detecting surface defects on parts, soldering defects, scratches on lithium batteries.
        
        .. image:: images/ano.png
            :scale: 100%

    - :ref:`Supervised Defect Segmentation` : Detects the **type** and **mask** of abnormal areas in an image. Use this model when you need to detect and classify abnormal areas in an image. Example scenarios: detecting multiple types of anomalies on part surfaces, ensuring components on a chip are installed correctly.
        
        .. image:: images/sem.png
            :scale: 100%

- Detecting text in images:

    - :ref:`OCR` : Detects **text** in images. Use this model when you need to extract text, letters, or symbols from an image.
        
        .. image:: images/ocr.png
            :scale: 100%