MSDA
A benchmark dataset and meta self-learning method for multi-source domain adaptation in scene text recognition.
What is MSDA?
Meta-SelfLearning is a research project that provides a benchmark dataset and method for multi-source domain adaptation in scene text recognition. It tackles the problem of domain shift by introducing a Meta Self-Learning approach, which combines pseudo-labeling with meta-learning to improve model performance across diverse text domains like car plates, documents, street signs, and handwritten text.
Researchers and practitioners in computer vision, particularly those working on domain adaptation, text recognition, or meta-learning. It is also relevant for developers building OCR systems that need to generalize across varied visual domains.
It offers the first multi-domain text recognition dataset with over five million images and a novel meta-learning-based method that outperforms traditional approaches in handling domain shift, providing a solid benchmark for future research.
Overview
Meta Self-learning for Multi-Source Domain Adaptation： A Benchmark
Use Cases
Best For
- Benchmarking multi-source domain adaptation methods for text recognition
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
