/Fine-Tune-Distillation

This work is the part of the environmental project during my time at Fujairah Research Center, Fujairah Environment Authority, UAE

Primary LanguageJupyter Notebook

Camel Farm Monitoring Framework

Overview

This repository contains the code for an automated framework designed for camel farm monitoring. The framework introduces two key contributions:

  1. Unified Auto-Annotation Framework:

    • Combines two models, GroundingDINO (GD), and Segment-Anything-Model (SAM), to automatically annotate raw datasets extracted from surveillance videos.

  2. Fine-Tune Distillation Framework:

    • Conducts fine-tuning of student models using the auto-annotated dataset.
    • Involves transferring knowledge from a large teacher model to a student model, resembling a variant of Knowledge Distillation.
    • Aims to be adaptable to specific use cases, enabling the transfer of knowledge from large models to small models, making it suitable for domain-specific applications.

Method Figure

Method Figure

Key Features

  • Automated Annotation: Utilizes GD and SAM models for automatic annotation of raw surveillance video datasets.

  • Fine-Tune Distillation: Conducts fine-tuning of student models for efficient real-time object detection.

  • Adaptability: Framework is designed to be adaptable to specific use cases, allowing knowledge transfer from large models to small models.

Dataset and Pre-trained Model

  • Raw Dataset: The framework leverages a raw dataset collected from Al-Marmoom Camel Farm in Dubai, UAE.

  • Pre-trained Model: GroundingDINO is used as the pre-trained teacher model while YOLOv8 as the student model for the Fine-Tune Distillation framework.

Deployable Model

The Fine-Tune Distillation framework produces a lightweight deployable model, YOLOv8, demonstrating high performance and computational efficiency for efficient real-time object detection.

Usage

To use this framework, follow the instructions provided in the corresponding directories for the Unified Auto-Annotation and Fine-Tune Distillation frameworks.

Citation

If you find this work useful for your research, please consider citing:

[Domain Adaptable Fine-Tune Distillation Framework For Advancing Farm Surveillance] [Raza Imam, Muhammad Huzaifa, Nabil Mansour, Shaher Bano Mirza, Fouad Lamghar] [Link will be out soon]