🔬 FEATURE GUIDE

Disclaimer: AI disease detection is an assistive tool. For critical or widespread crop disease, always consult your nearest Krishi Vigyan Kendra (KVK) or State Agriculture Department extension officer.

Crop diseases cause an estimated 20–40% of global crop yield loss each year. In India alone, diseases cost farmers over ₹50,000 crore annually in lost produce and treatment costs. THE FARMER'S CHALLENGE: - Most Indian farmers have no access to trained agronomists - Nearest Krishi Vigyan Kendra (KVK) may be 20–50 km away - By the time disease is identified through traditional channels, it has already spread to 30–50% of the field - Wrong diagnosis leads to wrong treatment — wasting money and damaging crops further THE KRISHOSAATHI SOLUTION: KrishoSaathi puts a trained AI agronomist in every farmer's pocket. By analysing a simple photo taken on any smartphone, KrishoSaathi can: - Identify the disease within 5–15 seconds - Assess severity (early / moderate / severe) - Recommend specific fungicides, pesticides, or cultural practices - Provide treatment instructions in the farmer's own language - Suggest when to consult a human expert

METHOD 1 — PHOTO UPLOAD (Recommended): STEP 1: Go to krishosaathi.in → Dashboard → "🔬 Disease Detection" section STEP 2: Tap "Upload Photo" STEP 3: Take a clear photo of the affected leaf, stem, or fruit (good lighting, close-up) STEP 4: Select your crop type from the dropdown (Rice, Wheat, Tomato, etc.) STEP 5: Tap "Analyse" STEP 6: Within 5–15 seconds, receive: • Disease name (common name + scientific name) • Severity level (Early / Moderate / Severe) • Cause (fungal / bacterial / viral / nutritional deficiency) • Immediate treatment steps • Preventive measures for the rest of the season • Whether to consult an expert METHOD 2 — TEXT SYMPTOM DESCRIPTION: If you cannot upload a photo (poor internet, no camera), describe your symptoms in text: - What part of the plant is affected? (leaves / stem / roots / fruit) - What does it look like? (yellow spots / brown patches / white powder / wilting / rotting) - How much of the field is affected? (few plants / 10% / 50%+) - When did you first notice it? KrishoSaathi's AI will analyse your description and provide a probable diagnosis.

KrishoSaathi currently supports disease detection for 13 major Indian crops: CEREALS & PULSES: - Rice — Blast, Brown Spot, Sheath Blight, Bacterial Leaf Blight, Tungro Virus, False Smut - Wheat — Rust (Yellow/Brown/Black), Powdery Mildew, Karnal Bunt, Loose Smut - Maize — Northern Corn Leaf Blight, Turcicum Leaf Blight, Downy Mildew - Tur/Arhar — Fusarium Wilt, Sterility Mosaic, Phytophthora Blight OILSEEDS: - Mustard — Alternaria Blight, White Rust, Downy Mildew, Sclerotinia Rot - Soybean — Soybean Mosaic Virus, Frogeye Leaf Spot, Phytophthora Root Rot VEGETABLES: - Tomato — Early Blight, Late Blight, Leaf Curl Virus, Bacterial Wilt, Fusarium Wilt - Potato — Late Blight, Early Blight, Black Scurf, Common Scab - Onion — Purple Blotch, Basal Rot, Downy Mildew FRUITS: - Banana — Sigatoka Leaf Spot, Panama Wilt, Bunchy Top Virus - Mango — Powdery Mildew, Anthracnose, Malformation Also supports: Cotton, Sugarcane, and general nutritional deficiency detection (N, P, K, Mg, Fe, Zn). More crops are being added with each update.

KrishoSaathi uses state-of-the-art Vision AI technology: PRIMARY MODEL: Groq Llama 4 Scout Vision (17B parameters) - Meta's latest multi-modal vision-language model - Specifically prompted for agricultural disease analysis - Runs on Groq's ultra-fast inference hardware (response in <5 seconds) FALLBACK MODEL: Groq Llama 3.2 Vision (11B parameters) - Activated if primary model is unavailable - Same agricultural prompting HOW THE AI ANALYSES YOUR PHOTO: 1. Your photo is uploaded to our secure server (Hetzner, Germany) 2. The photo is sent to Groq's API along with a detailed agricultural diagnostic prompt 3. The AI identifies visual patterns (colour, texture, shape, spread pattern) associated with known crop diseases 4. The AI cross-references with its training data of agricultural research papers, ICAR publications, and crop disease databases 5. Diagnosis, severity, and treatment are returned in your selected language 6. The photo is NOT permanently stored — it is deleted after analysis ACCURACY: - For clear, well-lit photos of distinctive diseases: 85–92% accuracy - For early-stage, atypical, or mixed infections: 65–75% accuracy - Text-based diagnosis (symptoms only): 70–80% accuracy The AI is trained primarily on common Indian crop diseases. Rare or region-specific diseases may not be accurately identified.

KrishoSaathi provides actionable treatment recommendations based on disease identification: IMMEDIATE ACTIONS: - Remove and destroy affected plant parts (isolation to prevent spread) - Specific fungicide/bactericide/insecticide recommendations with: - Chemical name (active ingredient) - Common brand names available in Indian market - Application rate (per acre or per litre of water) - Method of application (spray, soil drench, seed treatment) - Safety precautions (PPE, re-entry interval) PREVENTIVE MEASURES: - Crop rotation recommendations - Resistant variety suggestions - Optimal spacing and ventilation - Pre-season seed treatment protocols WHEN TO SEEK EXPERT HELP: KrishoSaathi will explicitly advise you to contact your KVK or State Agriculture Department when: - Disease severity is "Severe" (>30% crop affected) - Disease pattern suggests a new or unknown pathogen - Recommended treatment has not worked after 7 days - Disease outbreak affects multiple farmers in the village (possible epidemic) KVK HELPLINE: 1800-180-1551 (ICAR toll-free)

FOR ACCURATE DISEASE DETECTION, FOLLOW THESE TIPS: GOOD PHOTO PRACTICES: ✅ Take photo in natural daylight (avoid harsh midday shadows) ✅ Focus on the most clearly affected leaf or plant part ✅ Ensure the disease symptoms fill at least 40% of the frame ✅ Take multiple photos from different angles ✅ Include both affected and healthy tissue in the same frame for comparison ✅ File size: under 5MB (JPG, PNG, HEIC supported) AVOID: ❌ Blurry, out-of-focus photos ❌ Very dark or heavily shadowed images ❌ Photos taken from too far away (symptoms unclear) ❌ Heavily edited or filtered photos ❌ Photos of already dead/dried plant tissue (disease characteristics may be lost) ADDITIONAL CONTEXT HELPS: - Select the correct crop type from the dropdown - Mention crop growth stage (seedling / vegetative / flowering / grain filling) - Describe spread pattern (random plants / in rows / edge of field / entire field) More context = better AI diagnosis.