Skip to content
View AnastasiaLitvinova's full-sized avatar

Block or report AnastasiaLitvinova

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
AnastasiaLitvinova/README.md

Привет! 👋 Я Анастасия | Computer Vision Engineer

Специализация: Object Detection · Image Classification · Deep Learning

Создаю решения для компьютерного зрения в прикладных задачах, исследуя современные методы детекции огня и дыма на изображениях.


🛠️ Технологический стек

Основные инструменты

Python PyTorch OpenCV ResNet YOLO

Дополнительно

Docker Streamlit


🔥 Текущие проекты

  • Модель на основе ResNet18 с SE-блоками
  • Optuna для подбора гиперпараметров
  • Точность 92% с аугментацией данных
  • Демо-интерфейс на Streamlit
  • Исследую подходы:
    • YOLOv8 для object detection
    • Ансамбли с CNN-классификаторами (EfficientNet)
    • Оптимизация precision/recall для уменьшения false positives
  • Работа с датасетами:

🧪 Research Topics

  • Model Efficiency:
    • Quantization-Aware Training (QAT) схемы
    • Knowledge Distillation для YOLO → CNN ансамблей
  • Evaluation:
    • Анализ кривых Precision-Recall при class imbalance
    • Валидация на синтетических smoke-данных (GAN-генерация)

📐 Quality Framework

graph TD
    A[📤 Input Image] --> B[🦾 Model Inference]
    B --> C[📊 YOLO Evaluation]
    B --> D[⏱️ Inference Time]
    C --> E[🎚️ Threshold Tuning]
    D --> F[⚡ Hardware Optimization]
    
    style A fill:#5E35B1,stroke:#4527A0,color:#FFF
    style B fill:#3949AB,stroke:#283593,color:#FFF
    style C fill:#039BE5,stroke:#0288D1,color:#FFF
    style D fill:#00ACC1,stroke:#00838F,color:#FFF
    style E fill:#8FA94F,stroke:#2E7D32,color:#FFF
    style F fill:#FF9E0B,stroke:#FBC02D,color:#000
    
    classDef default font-family:Helvetica,font-size:14px
Loading

📫 Контакты

LinkedIn Kaggle Telegram Google


📈 GitHub статистика

GitHub Stats Top Langs

Popular repositories Loading

  1. 3DViewer_v2.0 3DViewer_v2.0 Public

    An application for viewing a 3D model created using C++, Qt, and the MVC pattern

    C++

  2. VisionForge VisionForge Public

    CIFAR-10 Image Classifier with PyTorch and Streamlit

    Jupyter Notebook

  3. LogInsight LogInsight Public

    Web server logs analyzer

    Python

  4. AnastasiaLitvinova AnastasiaLitvinova Public

  5. SmokeFinder SmokeFinder Public

    Detection of smoke and fire

    Jupyter Notebook