ConvNeXt Tiny (ImageNet-12k)
| Property | Value |
|---|---|
| Parameter Count | 36.9M |
| GMACs | 4.5 |
| License | Apache 2.0 |
| Framework | PyTorch (timm) |
| Paper | A ConvNet for the 2020s |
What is convnext_tiny.in12k?
ConvNeXt Tiny is a modern convolutional neural network designed for image classification tasks, trained on ImageNet-12k (a subset of ImageNet-22k with 11,821 classes). The model represents a modernized approach to ConvNet architecture, combining the best practices from transformers while maintaining the efficiency of traditional CNNs.
Implementation Details
This model features a carefully optimized architecture with 36.9M parameters and requires 4.5 GMACs of computation. It processes images at 224x224 resolution and generates 13.4M activations during inference. The model was trained using TPU infrastructure through the TRC program.
- Optimized for modern hardware with efficient compute requirements
- Implements state-of-the-art ConvNet architecture patterns
- Supports feature extraction and embedding generation
- Trained on a comprehensive dataset of 11,821 classes
Core Capabilities
- Image Classification with high accuracy
- Feature Map Extraction across multiple scales
- Image Embedding Generation
- Transfer Learning potential for downstream tasks
Frequently Asked Questions
Q: What makes this model unique?
This model combines modern architectural innovations with the efficiency of traditional CNNs, trained on a large-scale dataset (ImageNet-12k) while maintaining a relatively compact parameter count of 36.9M.
Q: What are the recommended use cases?
The model is well-suited for image classification tasks, feature extraction, and as a backbone for transfer learning in computer vision applications where moderate computational resources are available.
