ONGC PDF Image Extractor & Classifier

This project aims to automate the extraction of images from well completion report PDFs and classify them into predefined categories using machine learning techniques.

---

📌 Table of Contents

• Project Overview
• Features
• Usage
• Demo
• Report
• Future Improvements
• Acknowledgement
• License
• Contact

---

📌 Project Overview

This project automates the Extraction and Classification of Images from Well Construction Report PDFs using Machine Learning and Natural Language Processing (NLP) techniques.

The System:

• Detects figure with caption, figure without caption & graphs in PDFs using a YOLOv8 model.
• Extracts captions associated with detected images.
• Classifies captions using Logistic Regression-based NLP in classes:
  • Contour_Maps
  • Drilling_Plots
  • Geological_Map
  • Geotechnical_Order
  • Location_Map
  • Log_Motif
  • Remote_Sensing_Image
  • Seismic_Section
  • Stratigraphy_and_Casing_Plot
  • Structural_Map
  • Well_Construction_Diagram
  • Well_Schematic_Diagram
  • Others
• Organizes the output into structured directories.
• Provides a GUI-based interaction using PyQt6.

---

📌 Features

✅ Object Detection: Uses YOLOv8 to detect figures (labeled/unlabeled) and graphs.
✅ Caption Extraction: Extracts captions near detected images using PyMuPDF.
✅ NLP-based Classification: Classifies captions using TF-IDF + Logistic Regression.
✅ Automated Processing: Processes multiple PDFs at once.
✅ User-Friendly GUI: A PyQt6 interface for browsing PDFs and viewing results.
✅ Structured Output: Saves extracted images and captions in organized folders.

---

🚀 Usage

Prerequisites

• Python 3.6+

Clone the Repository

1. git clone https://github.com/mhsuhail00/ONGC-PDF-Image-Classification.git

2. cd ONGC-PDF-Image-Classification
   

Install Required Dependencies

pip install -r requirements.txt

Run the Application

python main.py

Output Directory

captured_images
   └───PDF_file_name
         ├───figure_without_label
         │             ├───page_1_object_2.png
         │             └───page_2_object_6.png
         ├───figure_with_label
         │             ├───Contour_Maps
         │                      ├───page_1_object_1.png
         │                      └───page_1_object_1.txt
         │             ├───Drilling_Plots
         │             ├───Geological_Map
         │             ├───Geotechnical_Order
         │             ├───Location_Map
         │             ├───Log_Motif
         │             ├───Others
         │             ├───Remote_Sensing_Image
         │             ├───Seismic_Section
         │             ├───Stratigraphy_and_Casing_Plot
         │             ├───Structural_Map
         │             ├───Well_Construction_Diagram
         │             └───Well_Schematic_Diagram
         └───graph
              ├───page_1_object_3.png
              └───page_1_object_4.png

---

📸 GIF Demonstration

---

📌 Future Improvements

• Enhancing Caption Extraction – Improving accuracy for multi-line captions.
• Deep Learning-Based Classification – Exploring transformer-based NLP models.
• Extending Image Classification – Using CNNs for better image categorization.
• GUI Enhancements – Adding real-time progress tracking.

---

🎯 Acknowledgments

This project was developed as part of an Industrial Training at ONGC GEOPIC Centre, Dehradun, under the guidance of Mr. Sanjay Chakravorty, Dy. General Manager (Programming), ONGC.

Author: Mohammad Suhail
Institution: Zakir Husain College of Engineering & Technology, Aligarh Muslim University

---

📄 Report

You can view the detailed report of this project here:
Project Report

---

📜 License

This project is licensed under the Apache License.

---

📩 Contact

• Developer: Mohammad Suhail
• Email: mhsuhail00@gmail.com
• GitHub Profile: mhsuhail00