Python Script for Extracting Diagrams from PDFs

**Objective:**
Create a robust, reusable Python script that automatically extracts proposed circuit diagrams (or other target figures) from research PDF documents, with minimal false positives (e.g., no text-only crops).

**Key Requirements:**
1. **Input:**
- A PDF file (e.g., `published_paper.pdf`).
- Ability to work with multi-page PDFs.

2. **Processing Steps:**
- Load and parse the PDF.
- Detect figures based on captions containing keywords (e.g., "Proposed", "Figure", "Circuit", "Diagram"). You need to identify which one is the proposed circuit/block diagrams , simulation waveforms etc
- Handle variable figure sizes—no fixed cropping.
- Support multiple figures per page.

3. **Output:**
- Save each detected figure as a separate PNG/JPG file.
- Save associated captions in `.txt` files.
- Output directory should be configurable.

4. **Technical Details:**
- Must work fully on **Windows**.
- Avoid dependency on manual PATH setup
- Use an AI model (Hugging Face, OpenAI, or equivalent) for caption detection if needed.
- Ensure no index errors or out-of-range issues.
- Must filter out crops that are mostly text.
- Support for batch processing multiple PDFs.
5. **Deliverables:**
- A single `.py` file, fully documented.
- Installation instructions (requirements.txt).
- Example run command and sample output for the provided test PDF.

6. **Bonus:**
- GUI or CLI options for keyword customization.


**Success Criteria:**
- Extracts only relevant figures with correct bounding boxes.
- Works out of the box with provided test PDFs for verification.
- No text in cropped images.