#Pet Image Quality Detection On-App AI

#Feature Introduction

This feature supports real-time quality inspection of images uploaded by users, automatically intercepting low-quality images such as blurry, distorted, and low resolution pictures to avoid ineffective transmission. Through local real-time filtering, it reduces dependency on cloud detection, significantly lowers network latency, enhances response efficiency, and provides a smoother upload experience for users.

#Process Description

1. Subject Recognition

   • Automatically identify the subject in the image (e.g., a pet) using object detection models
   • Output the subject category, position coordinates, and size information

2. Face Detection

   • Locate the pet's face area first
   • Identify key facial features (e.g., eyes, nose) afterward

3. Quality Verification

   • Brightness Detection: Calculate the average brightness of the subject area to filter out overexposed/underexposed images
   • Posture Detection: Calculate facial deflection angles based on feature points
   • Occlusion Detection: Determine if there are any obstructions to the face

#Interaction Flow

sequenceDiagram
    participant User
    participant Panel
    participant AI Engine
    participant Cloud

    rect rgb(200, 230, 255)
        User->>Panel: Enter the Create/Edit Pet Profile page
        Panel->>AI Engine: petsDetectCreate()
        AI Engine-->>Panel: Instance entity
    end

    rect rgb(200, 255, 225)
        Panel->>AI Engine: petsPictureQualityDetectForImage(images)
        AI Engine->>AI Engine: Analyze image quality
        alt High Quality
            AI Engine-->>Panel: {lowQuality: false}
        else Low Quality
            AI Engine-->>Panel: {lowQuality: true, reason: 11xx/12xx/13xx}
        end
    end

    rect rgb(255, 230, 200)
        alt Images meeting the standards are present
            Panel->>Cloud: Upload images that passed the detection
            Cloud-->>Panel: Return training results
            Panel->>User: Proceed to the next step
        else No images meeting the standards are present
            Panel->>User: Display error message
        end
    end

#Notes

1. It is recommended to reuse the same pet image quality detection instance within a single page.

2. Timely destruction of instances that are no longer in use is advisable.