tf_efficientnetv2_s.in21k_ft_in1k
| Property | Value |
|---|---|
| Parameter Count | 21.6M |
| Model Type | Image Classification |
| License | Apache-2.0 |
| Paper | EfficientNetV2: Smaller Models and Faster Training |
| Training Data | ImageNet-21k, Fine-tuned on ImageNet-1k |
What is tf_efficientnetv2_s.in21k_ft_in1k?
This is an EfficientNetV2-S model that represents a significant advancement in efficient deep learning architectures. Originally trained on ImageNet-21k and fine-tuned on ImageNet-1k, this PyTorch implementation was ported from the original TensorFlow model by Ross Wightman. The model demonstrates excellent balance between computational efficiency and accuracy, featuring 21.5M parameters and 5.4 GMACs.
Implementation Details
The model operates with F32 tensor type and is optimized for both training and inference. During training, it uses 300x300 image sizes, while for testing it employs 384x384 images. The architecture incorporates advanced features from the EfficientNetV2 family, offering improved training speed and model efficiency compared to its predecessors.
- Supports feature map extraction with multiple resolution outputs
- Provides image embedding capabilities with customizable output features
- Implements efficient image classification with top-k prediction support
Core Capabilities
- High-performance image classification on ImageNet-1k classes
- Feature extraction for downstream tasks
- Flexible image embedding generation
- Support for both training and inference workflows
Frequently Asked Questions
Q: What makes this model unique?
This model combines the benefits of training on ImageNet-21k with fine-tuning on ImageNet-1k, providing robust feature representation while maintaining efficiency with only 21.6M parameters. It represents an optimal balance between model size and performance.
Q: What are the recommended use cases?
The model is ideal for image classification tasks, feature extraction for transfer learning, and generating image embeddings for downstream applications. It's particularly well-suited for applications requiring a good balance between accuracy and computational efficiency.
