MixNet-L
| Property | Value |
|---|---|
| Parameter Count | 7.3M |
| Model Type | Image Classification |
| License | Apache-2.0 |
| Paper | MixConv: Mixed Depthwise Convolutional Kernels |
| Input Size | 224 x 224 |
What is mixnet_l.ft_in1k?
MixNet-L is an advanced image classification model that implements mixed depthwise convolutional kernels. This particular version has been fine-tuned on ImageNet-1k dataset, representing a carefully optimized variant of the original architecture. The model efficiently balances performance and computational resources with just 7.3M parameters while maintaining strong classification capabilities.
Implementation Details
The model architecture leverages mixed depthwise convolutions, operating at 0.6 GMACs with 10.8M activations. It's been specifically adapted from TensorFlow's "SAME" padding weights for PyTorch compatibility, making it versatile for various deep learning frameworks.
- Supports multiple usage modes: classification, feature extraction, and embedding generation
- Optimized for 224x224 pixel input images
- Implements efficient mixed kernel convolutions for better feature learning
Core Capabilities
- Image classification with 1000 classes (ImageNet-1k)
- Feature map extraction with multiple resolution outputs
- Generation of image embeddings for transfer learning
- Efficient inference with moderate computational requirements
Frequently Asked Questions
Q: What makes this model unique?
MixNet-L stands out for its use of mixed depthwise convolutional kernels, which combine different kernel sizes in a single operation. This approach provides better feature extraction capabilities while maintaining computational efficiency.
Q: What are the recommended use cases?
The model is well-suited for general image classification tasks, transfer learning applications, and as a feature extractor for downstream computer vision tasks. It's particularly effective when balanced performance and resource efficiency are required.
