Spaces:

pranit144
/

Crop_Recommendation_NPK

Sleeping

Parameter	Type	Required	Description	Example
nitrogen	float	Yes	Nitrogen content (kg/ha)	90
phosphorus	float	Yes	Phosphorus content (kg/ha)	42
potassium	float	Yes	Potassium content (kg/ha)	43
temperature	float	Yes	Average temperature (°C)	20.87
humidity	float	Yes	Relative humidity (%)	82.00
ph	float	Yes	Soil pH level (0-14)	6.50
rainfall	float	Yes	Average rainfall (mm)	202.93
location	string	Yes	Geographic location	"Maharashtra, India"

Example Request (curl)

curl -X POST http://localhost:7860/predict \
  -F "nitrogen=90" \
  -F "phosphorus=42" \
  -F "potassium=43" \
  -F "temperature=20.87" \
  -F "humidity=82.00" \
  -F "ph=6.50" \
  -F "rainfall=202.93" \
  -F "location=Maharashtra, India"

Example Request (Python)

import requests

data = {
    'nitrogen': 90,
    'phosphorus': 42,
    'potassium': 43,
    'temperature': 20.87,
    'humidity': 82.00,
    'ph': 6.50,
    'rainfall': 202.93,
    'location': 'Maharashtra, India'
}

response = requests.post('http://localhost:7860/predict', data=data)
print(response.json())

Example Request (JavaScript)

const formData = new FormData();
formData.append('nitrogen', '90');
formData.append('phosphorus', '42');
formData.append('potassium', '43');
formData.append('temperature', '20.87');
formData.append('humidity', '82.00');
formData.append('ph', '6.50');
formData.append('rainfall', '202.93');
formData.append('location', 'Maharashtra, India');

fetch('http://localhost:7860/predict', {
    method: 'POST',
    body: formData
})
.then(response => response.json())
.then(data => console.log(data));

Response (Success - 200)

{
  "predicted_crop": "RICE",
  "ai_suggestions": "RICE is an excellent choice for the given conditions with high humidity (82%) and substantial rainfall (202.93mm). The soil NPK values (90-42-43) are well-suited for rice cultivation, providing adequate nutrients for optimal growth.\n\nOther recommended crops:\n1. JUTE\n2. COCONUT\n3. PAPAYA\n4. BANANA",
  "location": "Maharashtra, India"
}

Response (Error - 500)

{
  "error": "An error occurred during prediction. Please try again.",
  "details": "Error message details"
}

AI Model Fallback Order (Text Generation)

NVIDIA Models (Phase 1):
- nvidia/llama-3.1-nemotron-70b-instruct
- meta/llama-3.1-405b-instruct
- meta/llama-3.1-70b-instruct
- mistralai/mixtral-8x7b-instruct-v0.1
Gemini Models (Phase 2):
- gemini-2.0-flash-exp
- gemini-1.5-flash
- gemini-1.5-flash-8b
- gemini-1.5-pro

🖼️ Endpoint 2: Image Analysis

`POST /analyze-image`

Analyzes agricultural images (crops, soil, plants) using AI vision models.

Request

Content-Type: multipart/form-data

Parameters:

Parameter	Type	Required	Description	Example
image	file	Yes	Image file (JPG, PNG)	crop_field.jpg
prompt	string	No	Custom analysis prompt	"Identify crop diseases"

Example Request (curl)

curl -X POST http://localhost:7860/analyze-image \
  -F "image=@/path/to/crop_image.jpg" \
  -F "prompt=Analyze this crop image and identify any diseases or issues"

Example Request (Python)

import requests

files = {
    'image': open('crop_image.jpg', 'rb')
}

data = {
    'prompt': 'Analyze this crop image and identify any diseases or issues'
}

response = requests.post('http://localhost:7860/analyze-image', 
                        files=files, 
                        data=data)
print(response.json())

Example Request (JavaScript)

const formData = new FormData();
const fileInput = document.querySelector('input[type="file"]');
formData.append('image', fileInput.files[0]);
formData.append('prompt', 'Analyze this crop image');

fetch('http://localhost:7860/analyze-image', {
    method: 'POST',
    body: formData
})
.then(response => response.json())
.then(data => console.log(data));

Response (Success - 200)

{
  "analysis": "The image shows a healthy rice crop in the vegetative stage. The plants display vibrant green color indicating good nitrogen availability. No visible signs of disease or pest damage. The crop appears to be well-watered with adequate spacing between plants. Recommendations: Continue current nutrient management, monitor for blast disease during heading stage, ensure proper water management.",
  "filename": "crop_image.jpg"
}

Response (Error - 400)

{
  "error": "No image file provided",
  "details": "Please upload an image file"
}

Response (Error - 500)

{
  "error": "An error occurred during image analysis. Please try again.",
  "details": "Error message details"
}

AI Model Fallback Order (Vision)

NVIDIA Vision Models (Phase 1):
- meta/llama-3.2-90b-vision-instruct
- meta/llama-3.2-11b-vision-instruct
- microsoft/phi-3-vision-128k-instruct
- nvidia/neva-22b
Gemini Vision Models (Phase 2):
- gemini-2.0-flash-exp
- gemini-1.5-flash
- gemini-1.5-flash-8b
- gemini-1.5-pro

Supported Image Formats

JPG/JPEG
PNG
WebP
BMP
GIF

Image Size Recommendations

Minimum: 224x224 pixels
Maximum: 4096x4096 pixels
File Size: < 10MB recommended

🏥 Endpoint 3: Health Check

`GET /health`

Returns system health status and configuration information.

Request

curl http://localhost:7860/health

Response (200)

{
  "status": "healthy",
  "nvidia_api_configured": true,
  "gemini_api_configured": true,
  "text_models_available": 8,
  "vision_models_available": 8
}

🔧 Configuration

Environment Variables

Create a .env file:

# Gemini API Key
GEMINI_API=your_gemini_api_key_here

# NVIDIA API Key
NVIDIA_API_KEY=your_nvidia_api_key_here

Model Configuration

Edit app.py to customize models:

# Text generation models
NVIDIA_TEXT_MODELS = [
    "nvidia/llama-3.1-nemotron-70b-instruct",
    # Add more models...
]

# Vision models
NVIDIA_VISION_MODELS = [
    "meta/llama-3.2-90b-vision-instruct",
    # Add more models...
]

# Gemini models (used for both text and vision)
GEMINI_MODELS = [
    "gemini-2.0-flash-exp",
    # Add more models...
]

🎯 Use Cases

Use Case 1: Crop Recommendation for Farmers

# Farmer inputs soil test results
data = {
    'nitrogen': 85,
    'phosphorus': 40,
    'potassium': 45,
    'temperature': 25.5,
    'humidity': 75,
    'ph': 6.8,
    'rainfall': 180,
    'location': 'Punjab, India'
}

response = requests.post('http://localhost:7860/predict', data=data)
result = response.json()

print(f"Recommended Crop: {result['predicted_crop']}")
print(f"AI Suggestions: {result['ai_suggestions']}")

Use Case 2: Disease Detection from Crop Images

# Upload crop image for disease analysis
files = {'image': open('diseased_crop.jpg', 'rb')}
data = {'prompt': 'Identify any diseases, pests, or nutrient deficiencies in this crop'}

response = requests.post('http://localhost:7860/analyze-image', 
                        files=files, data=data)
result = response.json()

print(f"Analysis: {result['analysis']}")

Use Case 3: Soil Quality Assessment

# Upload soil image for quality analysis
files = {'image': open('soil_sample.jpg', 'rb')}
data = {'prompt': 'Analyze soil quality, texture, and moisture content'}

response = requests.post('http://localhost:7860/analyze-image', 
                        files=files, data=data)
result = response.json()

print(f"Soil Analysis: {result['analysis']}")

🔄 Fallback System Behavior

Scenario 1: All APIs Working

Response Time: 1-3 seconds
Model Used: First NVIDIA model
Console Output: ✅ Success with first model

Scenario 2: NVIDIA API Down

Response Time: 3-6 seconds
Model Used: First available Gemini model
Console Output: Multiple ❌ for NVIDIA, then ✅ for Gemini

Scenario 3: All APIs Fail

Response Time: 8-15 seconds
Model Used: Generic fallback
Console Output: All ❌, generic response returned

📊 Response Times

Endpoint	Typical	With Fallback	Maximum
`/predict`	1-3s	3-6s	15s
`/analyze-image`	2-5s	5-10s	20s
`/health`	<100ms	N/A	<100ms

🐛 Error Handling

Common Errors

1. Missing Image File (400)

{
  "error": "No image file provided",
  "details": "Please upload an image file"
}

Solution: Ensure image field is included in multipart form data

2. Invalid Parameters (500)

{
  "error": "An error occurred during prediction. Please try again.",
  "details": "could not convert string to float: 'abc'"
}

Solution: Ensure all numeric parameters are valid numbers

3. All Models Failed (200 with fallback)

{
  "predicted_crop": "RICE",
  "ai_suggestions": "Note: AI suggestions are temporarily unavailable..."
}

Solution: Check API keys and internet connection

🧪 Testing

Test Script (Python)

import requests

# Test 1: Health Check
print("Testing health endpoint...")
health = requests.get('http://localhost:7860/health')
print(health.json())

# Test 2: Crop Prediction
print("\nTesting prediction endpoint...")
data = {
    'nitrogen': 90, 'phosphorus': 42, 'potassium': 43,
    'temperature': 20.87, 'humidity': 82.00, 'ph': 6.50,
    'rainfall': 202.93, 'location': 'Test Location'
}
prediction = requests.post('http://localhost:7860/predict', data=data)
print(prediction.json())

# Test 3: Image Analysis
print("\nTesting image analysis endpoint...")
files = {'image': open('test_image.jpg', 'rb')}
analysis = requests.post('http://localhost:7860/analyze-image', files=files)
print(analysis.json())

📚 Additional Resources

NVIDIA NIM API: https://docs.nvidia.com/nim/
Gemini API: https://ai.google.dev/docs
Flask Documentation: https://flask.palletsprojects.com/

🎉 Summary

Your Crop Recommendation System now has:

✅ 3 Powerful Endpoints

Crop prediction with AI suggestions
Image analysis for crops/soil
Health monitoring

✅ 16 AI Models Total

4 NVIDIA text models
4 NVIDIA vision models
4 Gemini models (text)
4 Gemini models (vision)

✅ Enterprise-Grade Reliability

Automatic failover
Graceful degradation
Comprehensive error handling

✅ Production Ready

RESTful API design
Proper error responses
Health check endpoint