Create app.py

app.py

@@ -0,0 +1,261 @@
+# app.py - PyIQA Image Quality Assessment API for Horsh
+import gradio as gr
+import pyiqa
+import torch
+from PIL import Image
+import requests
+from io import BytesIO
+import os
+
+# Загрузить модель при старте
+print("🚀 Loading ARNIQA model...")
+device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+metric = pyiqa.create_metric('arniqa', device=device)
+print(f"✅ Model loaded on {device}")
+
+def assess_image_quality(image):
+    """
+    Оценить качество изображения
+    Args:
+        image: PIL Image
+    Returns:
+        dict: {'quality': 0.75, 'score': 75, 'status': 'approved'}
+    """
+    try:
+        # Сохранить временно
+        temp_path = '/tmp/temp_image.jpg'
+        image.save(temp_path)
+
+        # Оценить качество (50-200ms)
+        with torch.no_grad():
+            score = metric(temp_path).item()
+
+        quality = score / 100.0  # Нормализовать 0-1
+        status = 'approved' if quality >= 0.3 else 'rejected'
+
+        return {
+            'quality': round(quality, 3),
+            'score': round(score, 2),
+            'status': status,
+            'threshold': 0.3,
+            'model': 'ARNIQA (WACV 2024)',
+            'device': str(device)
+        }
+
+    except Exception as e:
+        return {
+            'error': str(e),
+            'quality': 0.0,
+            'score': 0.0,
+            'status': 'error'
+        }
+
+def assess_from_url(url):
+    """Оценить по URL"""
+    try:
+        if not url or not url.startswith('http'):
+            return {'error': 'Invalid URL. Must start with http:// or https://'}
+
+        response = requests.get(url, timeout=15)
+        response.raise_for_status()
+
+        img = Image.open(BytesIO(response.content))
+
+        # Конвертировать в RGB если нужно
+        if img.mode != 'RGB':
+            img = img.convert('RGB')
+
+        return assess_image_quality(img)
+    except Exception as e:
+        return {
+            'error': f'Failed to load image: {str(e)}',
+            'quality': 0.0,
+            'score': 0.0,
+            'status': 'error'
+        }
+
+# Gradio Interface
+with gr.Blocks(
+    title="PyIQA Image Quality Assessment API",
+    theme=gr.themes.Soft()
+) as demo:
+
+    gr.Markdown("""
+    # 📸 Image Quality Assessment API
+
+    **Powered by ARNIQA** (WACV 2024) - State-of-the-art no-reference quality assessment
+
+    This API is used by [Horsh](https://hor.sh) photo-sharing app for automatic quality control.
+    """)
+
+    with gr.Tab("🖼️ Upload Image"):
+        with gr.Row():
+            with gr.Column():
+                image_input = gr.Image(type="pil", label="Upload Image")
+                upload_btn = gr.Button("Assess Quality", variant="primary", size="lg")
+
+                gr.Markdown("""
+                **Threshold:** Quality score ≥ 0.3 = Approved
+                **Model:** ARNIQA (Learning Distortion Manifold)
+                **Speed:** ~50-200ms per image
+                """)
+
+            with gr.Column():
+                upload_output = gr.JSON(label="Quality Assessment Result")
+
+        upload_btn.click(
+            fn=assess_image_quality,
+            inputs=image_input,
+            outputs=upload_output
+        )
+
+    with gr.Tab("🔗 Image URL"):
+        with gr.Row():
+            with gr.Column():
+                url_input = gr.Textbox(
+                    label="Image URL",
+                    placeholder="https://example.com/photo.jpg",
+                    lines=1
+                )
+                url_btn = gr.Button("Assess Quality", variant="primary", size="lg")
+
+                gr.Markdown("""
+                **Example URLs:**
+                - https://picsum.photos/800/600
+                - https://images.unsplash.com/photo-1506905925346-21bda4d32df4
+                """)
+
+            with gr.Column():
+                url_output = gr.JSON(label="Quality Assessment Result")
+
+        url_btn.click(
+            fn=assess_from_url,
+            inputs=url_input,
+            outputs=url_output
+        )
+
+    with gr.Tab("📚 API Documentation"):
+        gr.Markdown("""
+        ## REST API Usage
+
+        This Space exposes a REST API for programmatic access.
+
+        ### Python (gradio_client)
+
+        ```python
+        from gradio_client import Client
+
+        client = Client("YOUR_USERNAME/pyiqa-api")
+
+        # Assess from URL
+        result = client.predict(
+            "https://example.com/photo.jpg",
+            api_name="/assess_from_url"
+        )
+        print(result['quality'])  # 0.756
+        ```
+
+        ### Python (requests)
+
+        ```python
+        import requests
+        import json
+
+        url = "https://YOUR_USERNAME-pyiqa-api.hf.space/api/predict"
+
+        response = requests.post(url, json={
+            "data": ["https://example.com/photo.jpg"]
+        })
+
+        result = response.json()
+        quality = result['data'][0]['quality']
+        print(f"Quality: {quality}")
+        ```
+
+        ### Flutter/Dart
+
+        ```dart
+        import 'package:http/http.dart' as http;
+        import 'dart:convert';
+
+        Future<double> assessQuality(String imageUrl) async {
+          final response = await http.post(
+            Uri.parse('https://YOUR_USERNAME-pyiqa-api.hf.space/api/predict'),
+            headers: {'Content-Type': 'application/json'},
+            body: jsonEncode({'data': [imageUrl]}),
+          );
+
+          if (response.statusCode == 200) {
+            final result = jsonDecode(response.body);
+            return result['data'][0]['quality'];
+          }
+          throw Exception('Failed to assess quality');
+        }
+        ```
+
+        ### cURL
+
+        ```bash
+        curl -X POST https://YOUR_USERNAME-pyiqa-api.hf.space/api/predict \\
+          -H "Content-Type: application/json" \\
+          -d '{"data": ["https://example.com/photo.jpg"]}'
+        ```
+
+        ### JavaScript/TypeScript
+
+        ```javascript
+        const response = await fetch('https://YOUR_USERNAME-pyiqa-api.hf.space/api/predict', {
+          method: 'POST',
+          headers: { 'Content-Type': 'application/json' },
+          body: JSON.stringify({ data: ['https://example.com/photo.jpg'] })
+        });
+
+        const result = await response.json();
+        const quality = result.data[0].quality;
+        console.log('Quality:', quality);
+        ```
+
+        ## Response Format
+
+        ```json
+        {
+          "quality": 0.756,
+          "score": 75.6,
+          "status": "approved",
+          "threshold": 0.3,
+          "model": "ARNIQA (WACV 2024)",
+          "device": "cpu"
+        }
+        ```
+
+        ## Rate Limiting
+
+        - Free tier: No hard limits, but please be reasonable
+        - If you need high volume (>10k requests/day), contact us
+
+        ## Model Information
+
+        - **Paper:** [ARNIQA: Learning Distortion Manifold for Image Quality Assessment](https://arxiv.org/abs/2310.14918)
+        - **Conference:** WACV 2024 (Oral)
+        - **Code:** [miccunifi/ARNIQA](https://github.com/miccunifi/ARNIQA)
+        - **PyIQA:** [chaofengc/IQA-PyTorch](https://github.com/chaofengc/IQA-PyTorch)
+
+        ## About Horsh
+
+        This API powers quality control for [Horsh](https://hor.sh) - a photo-sharing app with AI-powered moderation.
+        """)
+
+    gr.Markdown("""
+    ---
+    **Note:** This is a public API. Please use responsibly. For production use, consider running your own instance.
+
+    **License:** Apache-2.0 | **Model:** ARNIQA (WACV 2024) | **Built with:** PyIQA + Gradio
+    """)
+
+# Запустить
+if __name__ == "__main__":
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=False
+    )