kahua-ml/nameplate-classifier

Nameplate Classifier

Model Description

This is a lightweight binary image classifier that determines whether an image contains a readable nameplate or not. The model is based on MobileNetV2 architecture and is optimized for industrial equipment nameplate detection.

Model Details

• Model Type: Image Classification (Binary)
• Architecture: MobileNetV2 backbone with custom classifier
• Framework: PyTorch
• Model Size: ~9.4MB
• Input Size: 224x224 RGB images
• Classes:
  • no_nameplate (0): Image does not contain a readable nameplate
  • has_nameplate (1): Image contains a readable nameplate

Training Data

The model was trained on the kahua-ml/nameplate-classification dataset, which contains:

• Positive examples: 3,456 images with nameplates from industrial equipment
• Negative examples: 826 images without nameplates from construction safety and industry datasets
• Total: 4,282 images

Performance

• Expected Accuracy: 85-95%
• Training Time: ~10-15 minutes
• Inference Speed: Very fast (optimized for real-time applications)

Usage

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

# Load model
model = torch.load("model.pth", map_location='cpu')
model.eval()

# Prepare image
transform = transforms.Compose([
    transforms.Resize((224, 224)),
    transforms.ToTensor(),
    transforms.Normalize(mean=[0.485, 0.456, 0.406], 
                        std=[0.229, 0.224, 0.225])
])

# Predict
image = Image.open("your_image.jpg")
input_tensor = transform(image).unsqueeze(0)

with torch.no_grad():
    outputs = model(input_tensor)
    probabilities = torch.nn.functional.softmax(outputs, dim=1)
    predicted = torch.max(outputs, 1)[1].item()
    confidence = probabilities[0][predicted].item()

result = "has_nameplate" if predicted == 1 else "no_nameplate"
print(f"Prediction: {result} (Confidence: {confidence:.3f})")

Applications

• Industrial equipment documentation
• Asset management systems
• Quality control in manufacturing
• Automated inventory systems

Limitations

• Optimized for industrial nameplates
• May not work well on severely damaged or obscured nameplates
• Requires clear, readable text on nameplates

Training Details

• Optimizer: Adam
• Loss Function: CrossEntropyLoss
• Epochs: 5
• Batch Size: 32
• Learning Rate: 0.001
• Data Augmentation: Random horizontal flip, rotation, color jitter

Model Architecture

LightweightNameplateClassifier(
  (backbone): MobileNetV2(
    (classifier): Sequential(
      (0): Dropout(p=0.2)
      (1): Linear(in_features=1280, out_features=128)
      (2): ReLU()
      (3): Dropout(p=0.3)
      (4): Linear(in_features=128, out_features=2)
    )
  )
)

Citation

If you use this model in your research, please cite:

@misc{kahua-nameplate-classifier,
  title={Lightweight Nameplate Classifier},
  author={Kahua ML Team},
  year={2024},
  howpublished={\url{https://huggingface.co/kahua-ml/nameplate-classifier}}
}