Text 2 Vector with GloVe

An AI system that converts text descriptions of shapes into vectors using GloVe 6B 100D embeddings and allows users to visualize the shape, saving the vector data to a JSON file. The shapes currently supported are circle, square, and triangle, and can be turned into vectors and plotted. letter A, letter B, and letter C can be plotted.

Developed as a proof-of-concept by Sen Dewael and Marnick Michielsen from Thomas More University (Geel, Belgium) in collaboration with KMITL (Bangkok, Thailand).

Project Status 🚧

• ✅ Text2Vec: Convert shape names into GloVe embeddings and visualize shapes
• ✅ Shape plotting functionality (circle, square, triangle, letter A, letter B and letter C)
• ⚙️ Future work: Add more shapes and functionalities

Note: This project is a proof-of-concept (POC). Some models or features may be incomplete or experimental.

---

Folder Structure 📂

root/
├── cli_text2shape.py   # CLI tool for interacting with the shapes and vectors
├── glove_loader.py     # Loads the GloVe embeddings for text-to-vector conversion
├── shape_library.py    # Contains shape plotting logic (circle, square, triangle)
├── viewer.py           # Loads and visualizes saved JSON vectors
├── saved_vectors/      # Folder where JSON files are saved

Features ✨

🖌️ Text2Vec: Convert shape names ("circle", "square", "triangle") to GloVe vectors and plot corresponding shapes

💾 Save vectors: Save the shape's name, vector values, and points to a JSON file in the saved_vectors/ directory

🔄 Visualization: Visualize saved vectors and shapes using viewer.py

✍️ Extensibility: More shapes can be added to shape_library.py for plotting

How It Works 🧠

Text2Vec Conversion: Convert shape names (e.g., "circle") to vectors using GloVe 6B 100D.

Shape Plotting: For recognized shapes like "circle", "square", or "triangle", the corresponding shape will be plotted.

Save and Load Vectors: After plotting, save the shape's name, vector, and plotted points to a JSON file in saved_vectors/. Use viewer.py to load and visualize saved vectors later.

Getting Started 🚀

Installation

Clone the repository:

git clone https://github.com/sendewael/text2vec.git
cd drawing-finisher

Install the required packages:

pip install numpy matplotlib

GloVe Embedding

This project uses the GloVe 6B 100D embeddings. You can download them from here. Once downloaded, unzip the file and place glove.6B.100d.txt in the project root directory.

Running the CLI (cli_text2shape.py)

To generate vectors for shapes and view them:

python cli_text2shape.py

The program will:

Ask you to input a shape name (e.g., circle, square, triangle).

Plot the shape corresponding to the name (if recognized).

Ask whether you'd like to save the shape's vector and plot data as a JSON file in the saved_vectors/ directory.

Loading and Visualizing Saved Vectors (viewer.py)

To visualize any saved shape vectors, run the viewer.py script:

python viewer.py

You will be prompted to enter the file path of the saved JSON vector file. The corresponding shape will be plotted.

Example Usage 🧑‍💻

Running the CLI and generating vectors:

$ python cli_text2shape.py
Enter a shape name (circle, square, triangle): circle

This will plot a circle and ask if you want to save the vector.

Visualizing a saved vector:

$ python viewer.py
Enter path to saved JSON vector file: saved_vectors/circle.json

This will load the saved vector and plot the corresponding circle.

Future Improvements 🔮

Add more shapes to shape_library.py (e.g., hexagon, pentagon, etc.)

Implement more sophisticated shape detection and vectorization

Develop a graphical interface (GUI or web-based)

Contributors 👥

Sen Dewael

Marnick Michielsen