EfficientNetV2-M Model
| Property | Value |
|---|---|
| Parameter Count | 54.4M |
| Model Type | Image Classification |
| License | Apache-2.0 |
| Paper | EfficientNetV2: Smaller Models and Faster Training |
| Training Data | ImageNet-21k + ImageNet-1k |
What is tf_efficientnetv2_m.in21k_ft_in1k?
This is an advanced implementation of EfficientNetV2-M, initially trained on ImageNet-21k and fine-tuned on ImageNet-1k. Originally developed in TensorFlow by the paper authors and successfully ported to PyTorch by Ross Wightman, it represents a significant advancement in efficient deep learning architectures.
Implementation Details
The model features 54.1M parameters and operates at 15.9 GMACs, with 57.5M activations. It's designed to process images at 384x384 during training and 480x480 during testing, showcasing impressive scalability and performance optimization.
- Specialized feature extraction capabilities
- Support for image embeddings generation
- Flexible architecture supporting both classification and backbone usage
- Optimized memory footprint with F32 tensor type
Core Capabilities
- High-accuracy image classification
- Feature map extraction across multiple scales
- Generation of image embeddings for downstream tasks
- Efficient processing of high-resolution images
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its efficient architecture that balances model size and performance, leveraging both ImageNet-21k pre-training and ImageNet-1k fine-tuning for enhanced feature representation.
Q: What are the recommended use cases?
The model excels in image classification tasks, feature extraction for downstream applications, and generating image embeddings for transfer learning scenarios. It's particularly suitable for applications requiring high-resolution image processing.
