Alice Benchmarks: Connecting Real World Object Re-Identification with the Synthetic.

Xiaoxiao Sun,Yue Yao,Shengjin Wang,Hongdong Li,Liang Zheng

DOI: https://doi.org/10.48550/arxiv.2310.04416

2023-01-01

Abstract:For object re-identification (re-ID), learning from synthetic data has becomea promising strategy to cheaply acquire large-scale annotated datasets andeffective models, with few privacy concerns. Many interesting research problemsarise from this strategy, e.g., how to reduce the domain gap between syntheticsource and real-world target. To facilitate developing more new approaches inlearning from synthetic data, we introduce the Alice benchmarks, large-scaledatasets providing benchmarks as well as evaluation protocols to the researchcommunity. Within the Alice benchmarks, two object re-ID tasks are offered:person and vehicle re-ID. We collected and annotated two challenging real-worldtarget datasets: AlicePerson and AliceVehicle, captured under variousilluminations, image resolutions, etc. As an important feature of our realtarget, the clusterability of its training set is not manually guaranteed tomake it closer to a real domain adaptation test scenario. Correspondingly, wereuse existing PersonX and VehicleX as synthetic source domains. The primarygoal is to train models from synthetic data that can work effectively in thereal world. In this paper, we detail the settings of Alice benchmarks, providean analysis of existing commonly-used domain adaptation methods, and discusssome interesting future directions. An online server has been set up for thecommunity to evaluate methods conveniently and fairly. Datasets and the onlineserver details are available at https://sites.google.com/view/alice-benchmarks.