| Property | Value |
|---|---|
| Parameters | 198M |
| GMACs | 34.4 |
| Training Image Size | 224x224 |
| Test Image Size | 288x288 |
| Top-1 Accuracy | 87.26% |
| Paper | ConvNeXt V2 Paper |
What is convnextv2_large.fcmae_ft_in22k_in1k?
This is a large-scale ConvNeXt V2 model that represents a significant advancement in convolutional neural network architecture. It was pretrained using a fully convolutional masked autoencoder (FCMAE) framework and subsequently fine-tuned on ImageNet-22k and ImageNet-1k datasets. The model demonstrates impressive performance with 87.26% top-1 accuracy on ImageNet-1k validation.
Implementation Details
The model architecture features 198 million parameters and requires 34.4 GMACs (billion multiply-accumulate operations) for inference. It processes images at 224x224 resolution during training and 288x288 during testing, with 43.1M activations during operation.
- Utilizes advanced FCMAE pretraining methodology
- Hierarchical feature extraction capabilities
- Optimized for both accuracy and computational efficiency
- Supports various input resolutions with adaptive pooling
Core Capabilities
- Image classification with 1000 classes
- Feature extraction for downstream tasks
- Generation of image embeddings
- Support for both inference and feature map extraction
Frequently Asked Questions
Q: What makes this model unique?
This model stands out due to its FCMAE pretraining approach and dual fine-tuning on ImageNet-22k and ImageNet-1k, resulting in robust feature representations. Its large-scale architecture with 198M parameters provides excellent performance while maintaining reasonable computational requirements.
Q: What are the recommended use cases?
The model excels in high-stakes image classification tasks, transfer learning applications, and as a feature extractor for computer vision tasks. It's particularly suitable for applications requiring high accuracy and robust feature representations, such as medical imaging, industrial inspection, and advanced computer vision systems.