fhf

readme.md

@@ -4,7 +4,7 @@ emoji: 🎨
 colorFrom: blue
 colorTo: purple
 sdk: gradio
-sdk_version: 4.44.0
+sdk_version: "4.44.0"
 app_file: app.py
 pinned: false
 license: mit
@@ -12,46 +12,30 @@ license: mit
 
 # 🎨 Intelligent Multi-Attribute Style Transfer
 
-An AI-powered system that analyzes images across multiple attributes and intelligently suggests relevant style transfers.
+This project intelligently analyzes image attributes to recommend and apply relevant style transfers. Instead of simply applying any filter, the system first classifies the image to ensure the suggested transformation makes contextual sense.
 
-## Features
+## Key Features
 
-- **Multi-Attribute Analysis**: Simultaneously classifies content, style, time of day, and weather
-- **Intelligent Recommendations**: Only suggests style transfers that make sense for your image
-- **User Choice**: Pick single effects or combine multiple transformations
-- **High Accuracy**: 90%+ accuracy on key classification tasks
+-   **Multi-Attribute Analysis**: Classifies images across four key attributes: content (human vs. landscape), style (photo vs. art), time of day, and weather.
+-   **Intelligent Recommendations**: Style transfers are only suggested when the AI's confidence in its analysis exceeds 60%, preventing irrelevant suggestions.
+-   **High-Accuracy Models**: Core classification models achieve 90%+ accuracy, ensuring reliable analysis.
+-   **Combined Effects**: Users can apply a single suggested effect or chain multiple transformations together.
 
-## How It Works
+## How to Use
 
-1. **Upload** your image
-2. **AI Analysis** across 4 different attributes
-3. **Smart Suggestions** based on confident predictions
-4. **Apply** single or combined style transfers
-
-## Available Transformations
-
-- 🌅 Day ↔ Night conversion
-- 🎨 Photo ↔ Japanese Art style
-- 🌫️ Fog removal (Foggy → Clear)
-- 🖼️ Content-aware enhancement
+1.  **Upload** an image.
+2.  **Review** the AI's analysis and smart style suggestions.
+3.  **Apply** your chosen effect(s).
 
 ## Technical Details
 
-This system uses multiple trained CNN models:
-
-- **Content**: Human vs Landscape classification (97% accuracy)
-- **Style**: Photograph vs Japanese Art classification (92% accuracy)
-- **Time**: Day vs Night classification (90% accuracy)
-- **Weather**: Foggy vs Clear classification (85% accuracy)
-
-Only confident predictions (>60%) trigger style transfer suggestions, ensuring relevant recommendations.
-
-## Model Architecture
+The system is powered by a series of fine-tuned CNN models built on ResNet50 and MobileNetV2 for high performance and accuracy.
 
-- Base: ResNet50 and MobileNetV2 with ImageNet pretraining
-- Training: Transfer learning with data augmentation
-- Deployment: TensorFlow models optimized for inference
+-   **Content Classification (Human/Landscape)**: 97% accuracy
+-   **Style Classification (Photo/Art)**: 92% accuracy
+-   **Time of Day Classification (Day/Night)**: 90% accuracy
+-   **Weather Classification (Foggy/Clear)**: 85% accuracy
 
 ---
 
-_Built with Gradio, TensorFlow, and ❤️_
+*Built with Gradio, TensorFlow, and ❤️*