Alice benchmarks

Alice benchmarks是一个包含AlicePerson和AliceVehicle两个数据集的系列，旨在促进合成数据与真实世界对象再识别的研究。AlicePerson和AliceVehicle分别针对人和车辆的再识别任务，数据收集自多种光照和图像分辨率条件，以模拟真实域适应测试场景。这些数据集的特点是训练集的聚类性未得到保证，更接近真实世界的数据分布。Alice benchmarks不仅提供了数据集，还包括评估协议和在线评估平台，以支持合成到真实域适应研究的发展。

Alice benchmarks is a series of datasets encompassing AlicePerson and AliceVehicle, aimed at promoting research on synthetic data and real-world object re-identification. AlicePerson and AliceVehicle target the person re-identification and vehicle re-identification tasks respectively, with data collected under diverse lighting conditions and image resolutions to simulate real-world domain adaptation test scenarios. A defining feature of these datasets is that the clustering property of their training sets is not guaranteed, rendering them more closely aligned with real-world data distributions. Beyond providing the datasets, Alice benchmarks also includes evaluation protocols and an online evaluation platform to support the advancement of synthetic-to-real domain adaptation research.