Grain Quality Classification Model 🌾

A multi-task deep learning model for grain quality control, trained on Tesla T4 GPU.

Model Description

This model performs multi-task learning for grain quality assessment:

Count Prediction: Estimates total grain count in images
Good Grain Count: Counts high-quality grains
Bad Grain Count: Counts low-quality/damaged grains
Quality Classification: Binary classification (good/bad dominant)

Architecture

Backbone: ResNet-50 (pre-trained on ImageNet)
Input Size: 256x256 RGB images
Multi-task heads: Separate heads for each prediction task
Training: Mixed precision on Tesla T4 GPU

Training Details

Epochs: 73
Best Validation Loss: 23.005850791931152
Optimizer: AdamW with OneCycleLR scheduler
Data Augmentation: Extensive augmentations for robustness
GPU: Tesla T4 with mixed precision training

Usage

import torch
from PIL import Image
import torchvision.transforms as transforms

# Load model
model = torch.load('pytorch_model.bin', map_location='cpu')
model.eval()

# Preprocessing
transform = transforms.Compose([
    transforms.Resize((256, 256)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

# Inference
image = Image.open('grain_image.jpg').convert('RGB')
input_tensor = transform(image).unsqueeze(0)

with torch.no_grad():
    outputs = model(input_tensor)
    
print(f"Count: {outputs['count'].item():.1f}")
print(f"Good grains: {outputs['good'].item():.1f}")
print(f"Bad grains: {outputs['bad'].item():.1f}")
print(f"Quality: {'Good' if outputs['quality'].argmax().item() == 1 else 'Bad'}")

Model Performance

Trained on agricultural grain dataset
Multi-task learning approach
Optimized for real-time quality control applications

Citation

If you use this model, please cite:

@misc{grain_classifier_2025,
  title={Grain Quality Classification Model},
  author={Your Name},
  year={2025},
  howpublished={\url{https://huggingface.co/Hk4crprasad/grain-quality}}
}

License

MIT License - See LICENSE file for details.