SMICA-ESA

---

🧪 What This Project Does

• ✅ Extracts 2D gnomonic sky patches from the B-mode polarization map.
• ✅ Trains a PyTorch autoencoder to detect structural anomalies.
• ✅ Computes a local variance map to detect statistical inhomogeneities.
• ✅ Compares B-mode anomalies with those from E-mode and T-mode.
• ✅ Saves anomaly patch images and analysis results for downstream use.

---

📡 Dataset

• Source: Planck Legacy Archive – SMICA CMB IQU maps
• File used: COM_CMB_IQU-smica_2048_R3.00_full.fits
• Resolution: NSIDE=2048 HEALPix format.

Ensure you have this file in data/ before running the notebooks.

---

🧠 Techniques Used

Task	Methods
Patch extraction	healpy.gnomview() (64×64 patches)
Preprocessing	MinMaxScaler, Flatten
Anomaly Detection	Autoencoder (PyTorch)
Statistical Anomaly	Variance map + μ + 2σ threshold
Visualization	Mollweide map, patch histograms
Optional	Isolation Forest / Clustering

---

🧰 Setup Instructions

1. Clone the repository:

git clone https://github.com/yourusername/cmb-bmode-anomaly.git
cd cmb-bmode-anomaly