# mnist-viz-specialist

> Use this skill to generate visual performance reports, confusion matrices, and sample digit galleries for MNIST models.

- Author: maymun maya
- Repository: maymun207/MNIST
- Version: 20260207053155
- Stars: 0
- Forks: 0
- Last Updated: 2026-02-07
- Source: https://github.com/maymun207/MNIST
- Web: https://mule.run/skillshub/@@maymun207/MNIST~mnist-viz-specialist:20260207053155

---

---
name: mnist-viz-specialist
description: Use this skill to generate visual performance reports, confusion matrices, and sample digit galleries for MNIST models.
---

# Mission
You are an expert at evaluating MNIST classification. When this skill is active:
1. Generate a Matplotlib grid showing 10 random test images, their true labels, and the model's predicted labels.
2. Color code the labels: Green for correct, Red for incorrect.
3. If the user is only using 3 digits (0, 1, 2), ensure the confusion matrix only shows those 3 classes.
4. Save the result as `performance_artifact.png` and present it as a Walkthrough Artifact.

# Constraints
- Always use 'Agg' backend for Matplotlib to ensure it runs in the agent's background terminal.
- Use `sns.heatmap` for the confusion matrix if `seaborn` is available.
---
name: mnist-web-dashboard
description: Builds a live Streamlit or Flask dashboard to visualize MNIST training and inference.
---

# Mission
When the user asks for a "live" or "web" dashboard:
1. Use **Streamlit** to create an interactive UI.
2. Include a "Canvas" component where the user can draw a digit (0, 1, or 2) and have the model predict it live.
3. Use the **Integrated Browser** (Port 8501) to display the UI.

# Execution Plan
- Check if `streamlit` is installed; if not, install it.
- Create `app.py` with a file uploader and a prediction result display.
- Run `streamlit run app.py --server.port 8501`.