selecsls42b.in1k
| Property | Value |
|---|---|
| Parameter Count | 32.5M |
| GMACs | 3.0 |
| Input Size | 224 x 224 |
| License | CC-BY-4.0 |
| Paper | XNect Paper |
What is selecsls42b.in1k?
selecsls42b.in1k is an efficient convolutional neural network designed for real-time computer vision tasks, particularly multi-person 3D motion capture. Developed as part of the XNect framework, this model balances computational efficiency with accuracy, featuring 32.5M parameters and operating on 224x224 pixel images.
Implementation Details
The model architecture is optimized for real-time performance while maintaining high accuracy. It utilizes selective sparse connections and achieves efficient feature extraction with just 3.0 GMACs (Giga Multiply-Accumulate Operations).
- Efficient backbone architecture with 4.6M activations
- Support for multiple output modes: classification, feature extraction, and embeddings
- Pretrained on ImageNet-1k dataset
- Implements both classification head and feature extraction capabilities
Core Capabilities
- Image Classification with 1000-class ImageNet support
- Feature map extraction with multiple resolution outputs
- Embedding generation for transfer learning tasks
- Real-time performance suitable for motion capture applications
Frequently Asked Questions
Q: What makes this model unique?
The model's architecture is specifically designed for real-time applications while maintaining good accuracy. Its selective sparse connections and efficient parameter usage make it particularly suitable for motion capture and real-time vision tasks.
Q: What are the recommended use cases?
This model is ideal for real-time applications including multi-person 3D motion capture, image classification tasks, and as a backbone for feature extraction in computer vision pipelines. It's particularly useful when deployment requires balancing performance with computational resources.
