This repository contains the codebase used by the SJTU-Thinklab-Det team for participating in DOTA-related competitions, which focus on object detection in aerial and remote-sensing imagery. The primary approach uses FPN-based two-stage detectors with ResNet backbones for both rotation and horizontal detection tasks. The project is connected to broader rotation detection benchmarks including MMRotate-Pytorch and AlphaRotate-TF, and reports performance metrics (mAP) on the DOTA1.0 dataset.