ReXNet-100
| Property | Value |
|---|---|
| Parameter Count | 4.84M |
| Model Type | Image Classification |
| License | MIT |
| Image Size | 224x224 |
| Top-1 Accuracy | 77.832% |
| Paper | Rethinking Channel Dimensions |
What is rexnet_100.nav_in1k?
ReXNet-100 is a lightweight convolutional neural network designed for efficient image classification. It's part of the ReXNet family, which introduces innovative channel dimension optimization techniques for better model efficiency. This particular variant is trained on ImageNet-1k and represents an excellent balance between model size and performance.
Implementation Details
The model utilizes a carefully designed architecture with 4.84M parameters and requires only 0.4 GMACs for inference. It processes images at 224x224 resolution with a crop percentage of 0.875, achieving 77.832% top-1 accuracy on ImageNet-1k.
- Optimized channel dimensions for efficient processing
- Balanced architecture with 7.4M activations
- Supports feature map extraction at multiple scales
- Implements efficient image embedding generation
Core Capabilities
- Image Classification with 1000 classes
- Feature Extraction with multiple feature map outputs
- Embedding Generation for downstream tasks
- Efficient inference with relatively low computational requirements
Frequently Asked Questions
Q: What makes this model unique?
ReXNet-100 stands out for its innovative approach to channel dimension optimization, providing an excellent balance between model size and accuracy. It's particularly suitable for resource-constrained applications while maintaining competitive performance.
Q: What are the recommended use cases?
This model is ideal for mobile and edge devices where computational resources are limited. It's well-suited for general image classification tasks, feature extraction, and as a backbone for transfer learning in computer vision applications.
