Domain generalisation for Object Detection

This repository is the collection of all the works done for the project on '[Domain Generalisation for Object Detection']. Here three popular object detectors (Faster R-CNN, FCOS, YoloV3) are explored and the codes related to each of the detector are present in their respective respositories.

Till date, the work done under this project has been published in two premier conference publications.

1. K. Seemakurthy, C. Fox, E. Aptoula, and P. Bosilj, "Domain Generalised Faster R-CNN", AAAI 2023 (We recommend to refer our updated 'repo' for replicating the results in this paper.)
2. K. Seemakurthy, P.Bosilj, E. Aptoula, and C. Fox, "Domain Generalised Fully Convolutional One Stage Object Detection", ICRA 2023 (Codes present in DGFCOS folder) The links for camera ready publications will be soon updated in this page.

However, the first version of ArXiv paper can be found at 'Domain Generalisation for Object Detection'.

We used the following datasets for our experiments.

1. Global Wheat Head Detection (GWHD)
2. Cityscapes
3. Sim10K
4. Berkely Deep Drive 100K (BDD100K)