tf_mobilenetv3_large_minimal_100.in1k
| Property | Value |
|---|---|
| Parameter Count | 3.95M |
| License | Apache-2.0 |
| Paper | Searching for MobileNetV3 |
| Architecture Type | MobileNetV3 Large Minimal |
| Task | Image Classification |
What is tf_mobilenetv3_large_minimal_100.in1k?
This is a highly efficient MobileNetV3 variant specifically designed for mobile and edge devices. Originally developed by TensorFlow researchers and later ported to PyTorch, it represents a minimal version of the MobileNetV3-Large architecture that maintains strong performance while reducing computational overhead. The model was trained on the ImageNet-1k dataset and achieves a good balance between accuracy and efficiency with just 3.95M parameters.
Implementation Details
The model utilizes an optimized architecture with 0.2 GMACs and 4.4M activations, designed to process 224x224 pixel images. It implements the MobileNetV3 architecture's key innovations while maintaining minimal computational requirements.
- Efficient feature extraction with progressive channel expansion
- Optimized for 224x224 input resolution
- Supports both classification and feature extraction modes
- Includes preprocessing transforms for input normalization
Core Capabilities
- Image Classification: Primary task with ImageNet-1k classes
- Feature Map Extraction: Supports multi-scale feature extraction
- Image Embeddings: Can generate fixed-size embeddings for transfer learning
- Mobile-Optimized: Designed for efficient mobile deployment
Frequently Asked Questions
Q: What makes this model unique?
This model represents a minimal version of MobileNetV3-Large, specifically optimized for efficiency while maintaining good accuracy. It's particularly notable for its small parameter count and efficient architecture design, making it ideal for mobile and edge deployments.
Q: What are the recommended use cases?
The model is best suited for mobile and edge device applications requiring image classification capabilities. It's particularly effective for scenarios where computational resources are limited but reasonable accuracy is still required. Common applications include mobile apps, edge devices, and real-time classification tasks.
