Test-time Adaptation

ImageNet-C is an open source data set that consists of algorithmically generated corruptions (blur, noise) applied to the ImageNet test-set. Source: Selective Brain Damage: Measuring the Disparate Impact of Model Pruning Image Source: https://arxiv.org/pdf/1807.01697.pdf

The DADE dataset, short for Driving Agents in Dynamic Environments, is a synthetic dataset designed for the training and evaluation of methods for the task of semantic segmentation in the context of autonomous driving agents navigating dynamic environments and weather conditions. This dataset was generated using the CARLA simulator (version 0.9.14) to provide perfect sensor synchronization and calibration, as well as precise semantic segmentation ground truths. All data were collected within the Town12 map in CARLA. DADE dataset is divided into two sub-datasets. For both subsets, each sequence is acquired by one agent (one ego vehicle) running for some time within a 5-hour time frame, amounting to a total of 990k frames for the entire dataset. The agents travel various locations such as forest, countryside, rural farmland, highway, low density residential, community buildings, and high density residential.