PerfSpectra is an interactive Streamlit-based tool for evaluating classification models. It helps researchers, data scientists, and ML practitioners analyze model performance with confusion matrices, classification reports, and mismatch analysis.
This tool is especially useful for writing shared task papers, as it automates error analysis and provides downloadable reports in multiple formats (CSV, PDF, TXT).
✅ Confusion Matrix 📊 – Visualizes model performance
✅ Classification Report 📝 – Precision, Recall, F1-score
✅ Mismatch Analysis 🔍 – Highlights misclassified samples
✅ Downloadable Reports 📥 – Get insights in CSV, PDF, and TXT
✅ Multi-file Support 📂 – Compare multiple prediction files
✅ Interactive UI 🎨 – Built with Streamlit for ease of use
1️⃣ Clone the repository:
git clone https://github.com/RJ-Hossan/PerfSpectra.git
cd PerfSpectra2️⃣ Install dependencies:
pip install -r requirements.txt3️⃣ Run the app:
streamlit run app.py1️⃣ Upload True Labels (CSV) with columns:
Id(Unique identifier)Label(True class labels)
2️⃣ Upload Prediction Files (CSV) with columns:
Id(Matching unique identifier)Label(Predicted class labels)
3️⃣ Get accuracy, confusion matrix, classification report, and mismatches
4️⃣ Download reports in CSV, PDF, or TXT format
Id and Label (case-insensitive)
This project is open-source, and contributions are welcome!
1️⃣ Fork the repo 🍴
2️⃣ Create a new branch 🔀
3️⃣ Make your changes ✨
4️⃣ Submit a pull request 📩
MIT License – Feel free to use and modify!
💬 Have suggestions? Want to contribute? Drop an issue or connect with me on LinkedIn
⭐ If you find this useful, don't forget to star the repo! 🌟