MobileNetV3 Small 100 LAMB
| Property | Value |
|---|---|
| Parameter Count | 2.55M |
| Model Type | Image Classification |
| License | Apache-2.0 |
| Paper | Searching for MobileNetV3 |
| Framework | PyTorch (timm) |
What is mobilenetv3_small_100.lamb_in1k?
This is a lightweight mobile-first image classification model based on the MobileNetV3 architecture, specifically optimized using the LAMB optimizer. It represents a careful balance between model efficiency and accuracy, designed for deployment on resource-constrained devices while maintaining competitive performance on ImageNet-1k classification tasks.
Implementation Details
The model is implemented using the timm library and trained with a sophisticated approach that includes LAMB optimization similar to the "ResNet Strikes Back" methodology, but with a 50% longer training duration and EMA weight averaging. It operates on 224x224 pixel images and utilizes F32 tensor types for computations.
- Employs LAMB optimizer with EMA weight averaging
- Features step-based learning rate scheduling with warmup
- Processes images at 224x224 resolution
- Achieves efficient inference with only 0.1 GMACs
- Maintains compact activation size of 1.4M
Core Capabilities
- Image classification with 1000 ImageNet classes
- Feature extraction for transfer learning
- Generation of image embeddings
- Mobile-optimized architecture for efficient deployment
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its optimization using the LAMB optimizer and its extremely efficient architecture with only 2.55M parameters, making it ideal for mobile and edge devices while maintaining good classification performance.
Q: What are the recommended use cases?
The model is best suited for mobile applications requiring image classification, feature extraction, or embedding generation where computational resources are limited but real-time performance is necessary.
