Diller / Languages:

• Türkçe
• English

Visual Analysis and OCR Processing Platform

An integrated platform for visual analysis using multiple AI providers (Google Gemini, OpenAI, Anthropic) and OCR/translation services.

🌟 Features

• Multi-Model Support: Gemini, OpenAI, Anthropic, HuggingFace
• OCR Integration: Text extraction with Google Vision API
• Translation: HF Models and Google Translate integration
• Cost Tracking: Real-time cost calculation per operation
• Log Management: Detailed logging and reporting system

⚙️ Setup

1. Requirements:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. API Keys: Create .env file:

GOOGLE_APPLICATION_CREDENTIALS="keys/google_service_account.json"
GOOGLE_TRANSLATE_API_KEY="your_google_translate_key"

3. Config File: Copy config example:

cp config.example.json config.json

🔧 Configuration

Main configuration file (config.json) structure:

{
  "models": {
    "gemini": {
      "enabled": true,
      "api_key": "${GEMINI_API_KEY}",
      "prompt_template": "Please analyze the image in detail:"
    },
    "openai": {
      "enabled": false,
      "api_key": "${OPENAI_API_KEY}",
      "endpoint": "https://api.openai.com/v1/chat/completions",
      "prompt_template": "Analyze the image according to the following criteria:"
    },
    "anthropic": {
      "enabled": true,
      "api_key": "",
      "endpoint": "https://api.anthropic.com/v1/messages",
      "prompt_template": "Please examine and analyze the image:"
    },
    "huggingface": {
      "enabled": true,
      "models": [
        {
          "name": "BLIP-Base",
          "model": "Salesforce/blip-image-captioning-base"
        }
      ]
    }
  },
  "ocr": {
    "google": {
      "enabled": true,
      "credentials": "path/to/google_service_account.json",
      "language_hints": ["en", "tr"]
    }
  },
  "translation": {
    "enabled": true,
    "hf_model": "Helsinki-NLP/opus-mt-en-tr"
  },
  "google_translate": {
    "enabled": false,
    "endpoint": "https://translation.googleapis.com/language/translate/v2"
  },
  "general": {
    "img_folder": "IMG",
    "results_folder": "results",
    "log_folder": "logs",
    "allowed_extensions": [".jpg", ".jpeg", ".png"]
  },
  "cost": {
    "gemini": {
      "variant": "gemini-1.5-flash",
      "default_variant": "gemini-1.5-flash",
      "variants": {
        "gemini-1.5-pro": {
          "token_threshold": 128000,
          "input_rate_low": 1.25,
          "input_rate_high": 2.50,
          "output_rate_low": 5.00,
          "output_rate_high": 10.00
        },
        "gemini-1.5-flash": {
          "token_threshold": 128000,
          "input_rate_low": 0.075,
          "input_rate_high": 0.15,
          "output_rate_low": 0.30,
          "output_rate_high": 0.60
        }
      }
    },
    "openai": {
      "divisor": 1000000,
      "variant": "gpt-4o",
      "rates": {
        "gpt-4o": {
          "input_rate": 5.00,
          "output_rate": 15.00
        }
      }
    },
    "anthropic": {
      "divisor": 1000000,
      "model_variant": "claude-3-haiku-20240307",
      "rates": {
        "claude-3-haiku-20240307": {
          "input_rate": 0.25,
          "output_rate": 1.25
        }
      }
    },
    "huggingface": 0.001,
    "ocr": {
      "free_units": 1000,
      "rate_low": 1.50,
      "rate_high": 0.60,
      "tier_threshold": 1000000,
      "units_per_page": 5
    },
    "translation": 0.0003
  }
}

🚀 Usage

python main.py

Image Requirements

• Place images in IMG/ folder
• Supported formats: JPG, PNG, JPEG

🔑 API Keys

1. Google Cloud:

   • Create service account via Google Cloud Console
   • Enable Vision API and Generative Language API

2. Anthropic:

   • Get API key from Anthropic Console

3. OpenAI (Optional):

   • Create key via OpenAI Platform

📊 Cost Management

Cost parameters in config.json:

"cost": {
  "gemini": {
    "variants": {
      "gemini-1.5-flash": {
        "input_rate_low": 0.075,
        "output_rate_low": 0.30
      }
    }
  },
  "anthropic": {
    "rates": {
      "claude-3-haiku": {
        "input_rate": 0.25,
        "output_rate": 1.25
      }
    }
  }
}

📂 File Structure

├── IMG/                   # Input images
├── results/               # Analysis results
├── logs/                  # System logs
├── config.json            # Main configuration
└── ocr_usage.json         # OCR usage tracking

🛠️ Troubleshooting

• OCR Errors: Verify Google service account path
• Model Loading Issues: Run pip install --upgrade transformers torch
• Permission Issues: Ensure .env file is in correct location

🤝 Contributing

To contribute:

1. Fork the repo
2. Create new branch: git checkout -b feature/new-feature
3. Commit changes: git commit -m 'Add some feature'
4. Push: git push origin feature/new-feature
5. Create Pull Request

🔓 Open Source

This project is developed as open source. You can freely use and modify it.

---

---

Görsel Analiz ve OCR Çözümleme Platformu

Çoklu AI sağlayıcıları (Google Gemini, OpenAI, Anthropic) ve OCR/Çeviri servisleri kullanarak görsel analiz yapan entegre bir platform.

🌟 Özellikler

• Çoklu Model Destek: Gemini, OpenAI, Anthropic ve HuggingFace
• OCR Entegrasyon: Google Vision API ile metin çıkarımı
• Çeviri: HF Modelleri ve Google Translate entegrasyonu
• Maliyet Takip: Her işlem için gerçek zamanlı maliyet hesaplama
• Log Yönetimi: Detaylı loglama ve raporlama sistemi

⚙️ Kurulum

1. Gereksinimler:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

2. API Anahtarları: .env dosyası oluşturun:

GOOGLE_APPLICATION_CREDENTIALS="keys/google_service_account.json"
GOOGLE_TRANSLATE_API_KEY="your_google_translate_key"

3. Config Dosyası: config.json örneğini kopyalayın:

cp config.example.json config.json

🔧 Yapılandırma

Ana yapılandırma dosyası (config.json) yapısı:

  "models": {
    "gemini": {
      "enabled": true,
      "api_key": "${GEMINI_API_KEY}",
      "prompt_template": "Please analyze the image in detail:"
    },
    "openai": {
      "enabled": false,
      "api_key": "${OPENAI_API_KEY}",
      "endpoint": "https://api.openai.com/v1/chat/completions",
      "prompt_template": "Analyze the image according to the following criteria:"
    },
    "anthropic": {
      "enabled": true,
      "api_key": "",
      "endpoint": "https://api.anthropic.com/v1/messages",
      "prompt_template": "Please examine and analyze the image:"
    },
    "huggingface": {
      "enabled": true,
      "models": [
        {
          "name": "BLIP-Base",
          "model": "Salesforce/blip-image-captioning-base"
        }
      ]
    }
  },
  "ocr": {
    "google": {
      "enabled": true,
      "credentials": "path/to/google_service_account.json",
      "language_hints": ["en", "tr"]
    }
  },
  "translation": {
    "enabled": true,
    "hf_model": "Helsinki-NLP/opus-mt-en-tr"
  },
  "google_translate": {
    "enabled": false,
    "endpoint": "https://translation.googleapis.com/language/translate/v2"
  },
  "general": {
    "img_folder": "IMG",
    "results_folder": "results",
    "log_folder": "logs",
    "allowed_extensions": [".jpg", ".jpeg", ".png"]
  },
  "cost": {
    "gemini": {
      "variant": "gemini-1.5-flash",
      "default_variant": "gemini-1.5-flash",
      "variants": {
        "gemini-1.5-pro": {
          "token_threshold": 128000,
          "input_rate_low": 1.25,
          "input_rate_high": 2.50,
          "output_rate_low": 5.00,
          "output_rate_high": 10.00
        },
        "gemini-1.5-flash": {
          "token_threshold": 128000,
          "input_rate_low": 0.075,
          "input_rate_high": 0.15,
          "output_rate_low": 0.30,
          "output_rate_high": 0.60
        }
      }
    },
    "openai": {
      "divisor": 1000000,
      "variant": "gpt-4o",
      "rates": {
        "gpt-4o": {
          "input_rate": 5.00,
          "output_rate": 15.00
        }
      }
    },
    "anthropic": {
      "divisor": 1000000,
      "model_variant": "claude-3-haiku-20240307",
      "rates": {
        "claude-3-haiku-20240307": {
          "input_rate": 0.25,
          "output_rate": 1.25
        }
      }
    },
    "huggingface": 0.001,
    "ocr": {
      "free_units": 1000,
      "rate_low": 1.50,
      "rate_high": 0.60,
      "tier_threshold": 1000000,
      "units_per_page": 5
    },
    "translation": 0.0003
  }
}

🚀 Kullanım

python main.py

Görsel Gereksinimleri

• Görseller IMG/ klasörüne yerleştirilmeli
• Desteklenen formatlar: JPG, PNG, JPEG

🔑 API Anahtarları

1. Google Cloud:

   • Google Cloud Console üzerinden servis hesabı oluşturun
   • Vision API ve Generative Language API etkinleştirin

2. Anthropic:

   • Anthropic Console üzerinden API anahtarı alın

3. OpenAI (Opsiyonel):

   • OpenAI Platform üzerinden anahtar oluşturun

📊 Maliyet Yönetimi

Maliyet parametreleri config.json içinde ayarlanır:

"cost": {
  "gemini": {
    "variants": {
      "gemini-1.5-flash": {
        "input_rate_low": 0.075,
        "output_rate_low": 0.30
      }
    }
  },
  "anthropic": {
    "rates": {
      "claude-3-haiku": {
        "input_rate": 0.25,
        "output_rate": 1.25
      }
    }
  }
}

📂 Dosya Yapısı

├── IMG/                   # Giriş görselleri
├── results/               # Analiz sonuçları
├── logs/                  # Sistem logları
├── config.json            # Ana yapılandırma
└── ocr_usage.json         # OCR kullanım takibi

🛠️ Sorun Giderme

• OCR Hatası: Google servis hesabı dosya yolunu kontrol edin
• Model Yükleme Hatası: pip install --upgrade transformers torch
• Yetki Sorunları: .env dosyasının doğru konumda olduğundan emin olun

🤝 Katkı

Katkıda bulunmak için:

1. Repoyu fork'layın
2. Yeni branch oluşturun: git checkout -b feature/new-feature
3. Değişiklikleri commit edin: git commit -m 'Add some feature'
4. Push işlemi: git push origin feature/new-feature
5. Pull Request oluşturun

📜 Lisans

MIT Lisansı - Detaylar için LICENSE dosyasına bakınız