gaohuang/MSDNet
A convolutional neural network architecture enabling anytime prediction and adaptive resource allocation for image classification.
Not currently ranked — collecting fresh signals.
star history
This repository implements MSDNet (Multi-Scale Dense Network), a deep learning architecture for efficient image classification. The network supports two inference modes: anytime prediction where classification confidence improves progressively, and batch-budget inference where computational resources are allocated dynamically based on example difficulty. The architecture uses multi-scale dense connections across classification layers to enable early exits for easier images.
Frequently asked
- What is gaohuang/MSDNet?
- A convolutional neural network architecture enabling anytime prediction and adaptive resource allocation for image classification.
- Is MSDNet open source?
- Yes — gaohuang/MSDNet is open source, released under the MIT license.
- What language is MSDNet written in?
- gaohuang/MSDNet is primarily written in Lua.
- How popular is MSDNet?
- gaohuang/MSDNet has 458 stars on GitHub.
- Where can I find MSDNet?
- gaohuang/MSDNet is on GitHub at https://github.com/gaohuang/MSDNet.