OVERVIEW
This chapter introduces you to the framework of YSAutoML, and provides links to detailed tutorials about YSAutoML
What is YSAutoML
YSAutoML is an open-source library designed for the automated construction of AI systems. It integrates independent utilities across data, network, and optimization into a unified platform, enabling efficient model building for tasks such as image recognition, segmentation, and object detection.
The project is developed by Yonsei CVLab, as part of a multi-year initiative on building a next-generation automated AI platform.
How to Use this Guide
Here is a detailed step-by-step guide to learn more about YSAutoML:
For installation instructions, please see get_started.
Major Features
Data Utilities
fyi: Dataset condensation (Flip Your Images), reducing large datasets into compact synthetic sets.
dsbn: Domain-Specific BatchNorm for source/target or augmentation-aware training.
Network Utilities (NAS)
fewshot: Few-shot NAS, exploring architectures with limited data.
zeroshot: Zero-shot NAS, searching architectures without labeled training data.
oneshot: One-shot NAS, training a supernet and searching subnets efficiently.
Optimization Utilities
fxp: Fixed-point quantization (W/A/G bitwidth 4–8).
losssearch: Automated loss function search.
mtlloss: Multi-task loss integration for classification, segmentation, and beyond.
Unified Workflow
Users specify task, dataset, resource budget, and memory constraints.
YSAutoML automatically builds and trains optimized AI systems under these requirements.
Library Structure
License
This project is released under the Apache 2.0 License
Last updated