PM Surya Ghar: Muft Bijli Yojana is a government scheme that aims to provide free electricity to households in India. The scheme was launched by Prime Minister Narendra Modi on February 15, 2024. Under the scheme, households will be provided with a subsidy to install solar panels on their roofs. The scheme is expected to benefit 1 crore households across India. It is estimated that the scheme will save the government Rs. 75,000 crore per year in electricity costs.

Governance for this scheme include the need to verify rooftop solar installations, so that subsidies reach genuine beneficiaries and public trust remains high. Field inspections alone are slow, costly, and uneven across states and DISCOMs. This challenge asks you to build a governance‑ready, auditable, and low‑cost remote verification digital pipeline that answers a simple question at any given coordinate (latitude, longitude):

“Has a rooftop solar system actually been installed here?”

Your system should work reliably across India's diverse roof types (sloped, flat roofs) and imaging conditions. The goal is not only accuracy, but also auditability and generalization across states.

Core Objectives (What your system must do)

When given an input folder location with a CSV containing list of geographic coordinates (latitude and longitude)

1. Fetch: Automatically retrieve or accept a recent, high-resolution rooftop image for a given (lat, lon). Be robust to small coordinate errors by searching within a configurable buffer (e.g., 10-20 m).

2. Classify: Run inference using your trained model and determine whether rooftop PV is present (binary: present / not present) and return a calibrated confidence score.

3. Quantify: If PV is present, estimate panel count, PV area (m²), and optionally capacity (kW) using a transparent assumption (e.g., Watt Peak/m²).

4. Explainability: Produce audit-friendly artifacts — polygon mask or bounding boxes, short “reason codes” (e.g., 'module grid', 'rectilinear array', 'racking shadows'), and a quality control (QC) status.

5. Store: JSON file for the test dataset into output folder location

Required QC status values:

· VERIFIABLE — Clear evidence either way (present/not present).

· NOT_VERIFIABLE — Insufficient evidence (e.g., low resolution, heavy shadow/cloud, occlusion by tanks/trees, stale imagery).

Ethics & Privacy Expectations

· Use only permissible imagery and respect licensing; cite all sources clearly.

· Document known biases (e.g., urban vs rural performance gaps) and mitigation steps.

Appendix: Definitions & Helpful Hints

· Capacity inference: document the Watt peak/m² or Watt peak/panel assumption with references.

· Buffer search: if (lat,lon) is slightly off, scan within ±10-20 m to lock onto the roof region.