Few Could Be Better Than All: Feature Sampling and Grouping for Scene Text Detection

Benchmark Model Rank Results
object-detection-in-aerial-images-on-dotaRBox–mAP: 79.59%