🎌 OtakuTag - Genre Prediction 🎌

OtakuTag is an AI-powered web application that predicts the genres of anime and manga based on their descriptions. Whether you're exploring new shows or want to know more about a manga, simply input a brief description, and OtakuTag will generate a list of genres associated with it.

---

📊 Features

• Multi-label Genre Classification: The model predicts multiple genres for a given anime/manga description.
• Data Scraping: Data is scraped from MyAnimeList to build a rich dataset for training the model.
• Data Cleaning: To ensure maximum utility of data.
• Model Training: We use distil-roberta-base for multi-label classification, incorporating:
  • Weighted Binary Cross-Entropy Loss for improved handling of imbalanced data.
  • Threshold Tuning to optimize for 13 distinct genres against F1 score for each.
  • Stratified Sampling to ensure a balanced dataset during model training.

---

🚀 How to Use

1. Visit the Web App: You can try out OtakuTag directly on Hugging Face Spaces:

   • Otaku-Tag on Hugging Face

2. Enter Description: Simply type in a short description of an anime or manga.

3. Get Predicted Genres: Click the "Get Predicted Genres" button, and you'll receive a list of relevant genres like "Slice of Life, Drama, Comedy, Action", and more!

---

🧠 Model Details

• Model: distil-roberta-base
• Loss Function: Weighted Binary Cross-Entropy
• Threshold Tuning: Optimized for predicting 13 genres.
• Sampling Strategy: Stratified sampling to ensure a balanced dataset for training.

---

🌍 Live Demo

• GitHub Webpage: Otaku-Tag GitHub Page

• Hugging Face Space: Otaku-Tag on Hugging Face

---

🛠️ Technologies Used

• Gradio: For the interactive front-end interface.
• Hugging Face Spaces: For easy model deployment and integration.
• Distil-Roberta-Base: A transformer-based model for multi-label text classification.
• Python: The primary language used for the backend logic and model training.

---

💻 Installation

To run the project locally, you can follow these steps:

1. Clone the repository:

   git clone https://github.com/your-repo/Otaku-Tag.git

2. Install Requirements:

   pip install -r requirements.txt

3. Fetch Models from my Hugginface repo and paste under the models folder :

   --Download Models From Here

4. Run the app.py :

   cd deployments
   python app.py