​

MobileNet

MobileNet is a lightweight computer vision model designed specifically for mobile and edge devices. It provides efficient image classification capabilities while maintaining a small footprint.

​

Overview

MobileNet is optimized for scenarios where computational resources are limited but real-time image classification is required. Key features include:

• Efficient architecture designed for mobile and edge devices
• Good balance between accuracy and model size
• Fast inference times
• Suitable for real-time applications

​

Usage

Here’s a simple example of how to use MobileNet for image classification:

from exla.models.mobilenet import MobileNet
from PIL import Image

# Initialize the model
model = MobileNet()

# Load an image
image = Image.open("path/to/your/image.jpg")

# Classify the image
predictions = model.classify(image)
print(predictions)

​

Example Output

The model returns a list of predictions with class labels and confidence scores:

[
    {"label": "golden retriever", "score": 0.85},
    {"label": "Labrador retriever", "score": 0.10},
    {"label": "tennis ball", "score": 0.03},
    # ... more predictions
]

​

Advanced Usage

​

Batch Processing

For processing multiple images efficiently:

images = [Image.open(f) for f in ["image1.jpg", "image2.jpg", "image3.jpg"]]
batch_predictions = model.classify_batch(images)

​

Custom Top-K

You can specify how many top predictions to return:

# Return only the top 3 predictions
predictions = model.classify(image, top_k=3)

​

Performance Considerations

MobileNet is designed for efficiency, but there are still some considerations for optimal performance:

• Memory usage: ~5-10MB
• Inference time: Typically 10-50ms on modern devices
• Power consumption: Lower than larger models like ResNet

​

Example Applications

• Real-time object recognition in mobile apps
• Smart camera features
• Augmented reality applications
• IoT devices with visual recognition capabilities

​

Limitations

• Lower accuracy compared to larger models like ResNet
• Limited ability to detect small objects or fine details
• Performance varies based on image quality and lighting conditions

​

Comparison with Other Models

For applications requiring higher accuracy and where computational resources are less constrained, consider using ResNet34 instead.

For more information on optimizing model performance, see the Custom Models Optimization guide.