AI proposal analyzer
Automated quality evaluation across seven criteria. This is decision support for faculty — the final selection is made by the panel in the offline round.
Demo data
KrishiScan — Offline AI Crop Disease Diagnosis
Team Innovexa · SIH1601
86Overall
Analysed 28 Jul 2026, 10:20 am
ShortlistedEvaluation breakdown
Problem Understanding90/100
Innovation84/100
Feasibility88/100
Technical Strength87/100
Impact89/100
Scalability82/100
Clarity84/100
AI summary
A technically credible and well-scoped agricultural AI proposal with a genuine differentiator in offline on-device inference. The main gap is evaluation rigour — the team should quantify model performance and address misdiagnosis risk before the presentation round.
Strengths
3- Clear articulation of the offline-first constraint and a concrete technical answer to it (quantised 6.8 MB model).
- Strong dataset strategy — locally collected Pune-district images meaningfully differentiate this from generic PlantVillage solutions.
- Measurable impact claim tied to a specific farm size rather than vague social benefit.
Weaknesses
3- Model accuracy figures for the locally collected dataset are not reported.
- Voice output layer is described but no language coverage or TTS approach is specified.
- No discussion of what happens when the model is uncertain — false diagnosis risk is unaddressed.
Recommendations
3- Add a confidence threshold with a 'consult KVK officer' fallback for low-confidence predictions.
- Report top-1 / top-5 accuracy and confusion between visually similar diseases.
- Include a cost-of-deployment estimate per 1,000 farmers to strengthen the scalability section.
Missing information
3- Model accuracy metrics
- Risk & mitigation table
- Data privacy note for uploaded farm images