## Pet Image Quality Detection Solution <span class="tag_h2">On-App AI</span>

### Pain Point Analysis

To ensure pet smart devices (such as feeders) can provide accurate services, establishing high-quality pet profiles is crucial. However, during actual profile creation, users frequently upload images with quality issues, which leads to:

- Decreased accuracy of cloud AI model recognition
- Delayed response in cloud AI model detection
- Wasted network transmission resources
- Invalid token consumption

<div style="display: flex; align-items: center; justify-content: flex-start; margin-top: 10px;">
  <div style="text-align: center; width: 200px; margin-right: 5px">
    <img alt="Original video" src="https://images.tuyacn.com/rms-static/ab56fd70-5651-11f0-9fb9-e1834df84344-1751356921543.jpg?tyName=image_1.jpg" width="200">
    <p style="margin: 0;">(Subject obstructed)</p>
  </div>
  <div style="text-align: center; width: 200px; margin-right: 5px">
    <img alt="Original video" src="https://images.tuyacn.com/rms-static/1689e4b0-5655-11f0-9fb9-e1834df84344-1751358389883.jpg?tyName=image_4.jpg" width="200">
    <p style="margin: 0;">(Abnormal facial angle)</p>
  </div>
  <div style="text-align: center; width: 200px; margin-right: 5px">
    <img alt="Original video" src="https://images.tuyacn.com/rms-static/c9590700-5651-11f0-9fb9-e1834df84344-1751356971888.jpg?tyName=image_2.jpg" width="200">
    <p style="margin: 0;">(Subject too small)</p>
  </div>
  <div style="text-align: center; width: 200px;">
    <img alt="Original video" src="https://images.tuyacn.com/rms-static/0bfd8ba0-5655-11f0-9fb9-e1834df84344-1751358372186.jpg?tyName=image_3.jpg" width="200">
    <p style="margin: 0;">(Overexposed image)</p>
  </div>
</div>

### Solution

<img alt="Original video" src="https://images.tuyacn.com/rms-static/83856d10-564f-11f0-9fb9-e1834df84344-1751355995745.jpg?tyName=select_image_1.jpg" width="250" style="margin-top: 8px">

**On-Device Real-Time Detection**

- Local processing on mobile devices
- Quality assessment completed within 1 second
- Immediate rejection of substandard images

**Triple Quality Assurance**

| **Detection Type**        | **Inspection Items**     | **Standards/Requirements**   |
| ------------------------- | ------------------------ | ---------------------------- |
| **Basic Detection**       | Image clarity/brightness | Meets recognizable standards |
| **Composition Detection** | Single pet verification  | Contains only one pet        |
|                           | Subject proportion       | ≥30%                         |
| **Pose Detection**        | Front face angle         | ≤45°                         |
|                           | Side face angle          | ≤40°                         |

### Technical Advantages

| **Technical Features**               | **Functional Characteristics**     | **Detailed Description**                             |
| ------------------------------------ | ---------------------------------- | ---------------------------------------------------- |
| **Lightweight Dynamic Architecture** | 10MB-level lightweight model       | Minimal resource consumption                         |
|                                      | Supports hot model updates         | Model updates without redeployment                   |
|                                      | On-demand loading mechanism        | Dynamically loads required modules, saving resources |
| **Intelligent Filtering System**     | Multi-pet/foreign object filtering | Automatically excludes interference elements          |
|                                      | Privacy face filtering             | Detects and blurs human faces                        |
|                                      | Device adaptive processing         | Adapts to different hardware capabilities            |
| **Privacy Security**                 | Local data processing              | All computations performed on device                 |
|                                      | No user photo uploads              | Eliminates privacy data leakage risks                |
|                                      | Complies with privacy regulations  | Meets GDPR and other international standards         |

### Application Scenarios

| Application Scenario   | Functional Description                                                             |
| ---------------------- | ---------------------------------------------------------------------------------- |
| Digital Pet Profiles   | Establish complete digital identity information for pets, recording growth history |
| Smart Feeding System   | Record multi-pet feeding situations through image recognition                      |
| Pet Insurance Services | Verify pet identity authenticity and confirm injury/illness during claims          |