• 

  “Hauntify in action: Predicting spooky vibes and recommending the perfect Halloween costume with just a click!”

Hauntify 🎃

Try it out

• 🔗 Try out the full interactive Hauntify Halloween Fun experience here!: Click to launch Hauntify 🎃
• 👩‍💻 Explore the complete Hauntify project code, datasets, and instructions on GitHub: Hauntify GitHub Repository

---

Inspiration

Halloween has always been something I look forward to — from haunted houses to creative costumes. I wanted to build something that captured the spooky vibe while using what I’ve learned in machine learning. That’s how Hauntify came to life — a fun little project that mixes spooky fun with data science and creativity.

---

What it does

Hauntify has two main features:

It takes a sentence and predicts how spooky it sounds using a machine learning model.

It recommends a Halloween costume based on your selected mood and favorite genre. It’s simple, interactive, and made to bring a Halloween twist to data science!

---

How we built it

I created a small dataset of Halloween-themed sentences labeled as spooky or not. Using TF-IDF Vectorization and Logistic Regression, I trained a model to detect spooky words and phrases. For the costume recommender, I used a rule-based system with a CSV file that matches mood and genre with the best costume. Finally, I built the entire interface using Streamlit, added emojis, markdown, and a few layout tweaks to make it fun to use.

---

Challenges We Ran Into

• Gathering a quality dataset of spooky sentences.
  
• Ensuring the ML model generalized well to unseen sentences.
  
• Balancing fun and functional design in Streamlit.
  

---

Accomplishments

• Built a fully interactive app in under a week.
  
• Achieved ~92% accuracy on the spooky sentence classifier.
  
• Developed a creative rule-based costume recommender.
  

---

What We Learned

• Practical TF-IDF + Logistic Regression applications for text classification.
  
• Streamlit for interactive web apps with Python.
  
• Balancing ML logic with user-friendly design.
  

---

What's Next for Hauntify

• Expand the dataset to improve accuracy.
  
• Add more costume options and mood/genre mapping.
  
• Include spooky image filters and more interactive features.
  

---

Built With

• Python 3.13
  
• Streamlit
  
• OpenCV
  
• NumPy
  
• Pillow
  
• scikit-learn
  
• Pandas
  

---

Contributors

• Suwaasha Murugaperumal
• Sivaksha Sivagami Arumugavelu Palanidevi

Built With

• csv
• jupyter
• logistic-regression
• machine-learning
• numpy
• opencv
• pandas
• pil)
• pillow
• rgb
• scikit-learn
• streamlit
• technologies:-python
• tf-idf
• training