Pokémon Multilabel Classification

This is a Machine Learning project focused on multilabel classification of Pokémon. The goal is to predict types for each Pokémon using their images. The project includes data preprocessing, model training, evaluation, and documentation.

---

📁 Project Structure

• compiled_data/: Contains the image paths and their corresponding type relations.
• data/: Includes the Pokémon sprites and their labels, separated into train, validation, and test datasets.
• evaluation/: Script for evaluating Pokémon images against predicted types.
• pokedata.py: Script for data preprocessing, ensuring the data is ready for model training.
• Final report/: The final report for the Machine Learning course (in Portuguese).
• Images/: Visual resources related to the models and dataset.
• Models/: Contains model scripts, including:
  • tds.py
  • tds Da.py
  • MobileNet.py
• PokeAPI/: Data fetched from PokeAPI. Huge thanks to them for providing publicly available Pokémon data!
• Trained models/: Pre-trained model files in .h5 format.

---

🛠️ How to Use

1. Set Up Environment:

   • Install the required Python libraries installed for Machine Learning and data processing.

2. Prepare the Dataset:

   • Use the scripts in pokedata.py to preprocess the data if needed.
   • Verify the data structure in the data/ directory (organized into train, validation, and test).

3. Train a Model:

   • Run one of the model scripts (tds.py, tds Da.py, or MobileNet.py) to train your model on the provided dataset.

4. Evaluate Results:

   • Use the evaluation script in evaluation/ to test the model and compare predictions against true labels.

---

🌟 Acknowledgments

• PokeAPI: Special thanks for making the Pokémon data publicly accessible. Visit PokeAPI for more information.
• This project was completed as part of the Machine Learning course requirements.

---

📄 Final Report

The final report, written in Portuguese, can be found in the Final report/ directory. It includes detailed explanations of the project objectives, methods, and outcomes.

---

📫 Contact

If you have questions or suggestions, feel free to reach out via GitHub Issues.