This is the unsupervised domain adaptation object detection implementation based on the MMDetection .
Until now, we have reproduced some SOTA UDAOD models including DAF, MAF, and DeepAlign on our custom datasets. In the future, these models will be valided on the CityScapes and FoggyCityScapes.
Install Pytorch=1.7.1 or 1.8.0, mmcv-full=1.3.17, and mmdetection=2.19 as official guidiance. Then, install the mmdet as our repo again.
cd Unsupervised_Domain_Adaptation_Object_Detection_Implementation
python setup.py installI'm preparing my dissertation and defense recently. Guidelines and public dataset validation will be conducted after I complete my dissertation.