Quality assessment and grading of onions are often subjective and vary across procurement centers, resulting in disputes and inconsistencies.
“Onion quality assessment during procurement is currently done manually, leading to heavy human bias, disputes between farmers and buyers, and inconsistent grading. We need to build a smart mobile application that scans batches of onions using a smartphone camera to automatically detect defects, calculate grades, and generate an instant digital report.”
Deconstructed Problem Blueprint
WHAT TO BUILD?
An AI-powered mobile application (cross-platform using Flutter/React Native) integrated with computer vision models for real-time object detection and classification of agricultural produce.
WHO IS IT FOR?
Problem owners are the Department of Consumer Affairs (DoCA) and procurement agencies (like NAFED/NCCF). Users include procurement officers, mandi inspectors, farmers, and cooperative society managers.
WHY DOES IT EXIST?
Manual inspection is subjective, slow, and prone to corruption or human fatigue. Different procurement centers follow varying standards, creating massive supply chain friction, pricing disputes, and post-harvest losses.
INPUT → PROCESS → OUTPUT ARCHITECTURE PIPELINE
Smartphone camera images or short video feeds of onion heaps/crates, metadata (batch ID, procurement center location, timestamp, farmer ID).
On-device or edge-cloud AI inference using YOLO-based object detection to count onions, segment sizes, and classify surface defects (rotted, sprouted, damaged, skin peeling, black mold). Post-processing algorithms calculate the ratio of Grade A versus URS (Unspecified/Reject Standard) onions.
An automated digital quality certificate with breakdown percentages, visual heatmaps of defects on the scanned batch, a unique QR code for traceability, and cloud synchronization for audit logs.
CONSTRAINTS
- Must work offline-first in rural mandis with poor or no internet connectivity.
- Inference must run fast on mid-range Android smartphones commonly used by field officers.
- Lighting variations in open-air mandis require robust color correction and normalization pipelines.
- Must comply with standard agricultural grading parameters specified by the government.
JURY METRICS
- Defect detection accuracy (mAP > 85% for rot, sprouts, and sizing).
- Processing latency under 3 seconds per batch image.
- UI/UX simplicity score (usable by non-tech-savvy workers in 2 taps).
- Reliability of offline data sync when network reconnects.
CORE BOTTLENECK
Handling overlapping onions in a heap and accurately estimating subsurface or hidden rot without manual sorting.
Why You Should Choose This PS (Pros)
- Extremely high social and economic impact aligned directly with government priorities (Digital Agriculture & Fair Trade).
- Clear, self-contained scope with concrete inputs and outputs that are easy to demo live.
- Strong 'wow factor' during evaluation when judges see instant AI grading live through a phone camera.
Challenges & Risks to Watch (Cons)
- Lack of a standardized public dataset means teams must curate and annotate their own onion image dataset during the hackathon.
- Complex edge cases involving lighting, camera angles, and occlusions in real-world onion bags.
Recommended Tech Stack & Grand Finale Features
Killer Features to Impress the SIH Jury
View Official SIH 2026 Portal Raw Text▼
Develop an AI-based mobile application that: