Test datasets should contain many different challenging aspects so that the robustness and real-world applicability of algorithms can be assessed. In this work, we present a new test dataset for semantic and instance segmentation for the automotive domain. We have conducted a thorough risk analysis to identify situations and aspects that can reduce the output performance for these tasks. Based on this analysis we have designed our new dataset. Meta-information is supplied to mark which individual visual hazards are present in each test case. Furthermore, a new benchmark evaluation method is presented that uses the meta-information to calculate the robustness of a given algorithm with respect to the individual hazards. We show how this new approach allows for a more expressive characterization of algorithm robustness by comparing three baseline algorithms.

In recent years, a great number of datasets were published to train and evaluate computer vision (CV) algorithms. These valuable contributions helped to push CV solutions to a level where they can be used for safety-relevant applications, such as autonomous driving. However, major questions concerning quality and usefulness of test data for CV evaluation are still unanswered. Researchers and engineers try to cover all test cases by using as much test data as possible. In this paper, we propose a different solution for this challenge. We introduce a method for dataset analysis which builds upon an improved version of the CV-HAZOP checklist, a list of potential hazards within the CV domain. Picking stereo vision as an example, we provide an extensive survey of 28 datasets covering the last two decades. We create a tailored checklist and apply it to the datasets Middlebury, KITTI, Sintel, Freiburg, and HCI to present a thorough characterization and quantitative comparison. We confirm the usability of our checklist for identification of challenging stereo situations by applying nine state-of-the-art stereo matching algorithms on the analyzed datasets, showing that hazard frames correlate with difficult frames. We show that challenging datasets still allow a meaningful algorithm evaluation even for small subsets. Finally, we provide a list of missing test cases that are still not covered by current datasets as inspiration for researchers who want to participate in future dataset creation.

Performance characterization of stereo methods is mandatory to decide which algorithm is useful for which application. Prevalent benchmarks mainly use the root mean squared error (RMS) with respect to ground truth disparity maps to quantify algorithm performance. We show that the RMS is of limited expressiveness for algorithm selection and introduce the HCI Stereo Metrics. These metrics assess stereo results by harnessing three semantic cues: depth discontinuities, planar surfaces, and fine geometric structures. For each cue, we extract the relevant set of pixels from existing ground truth. We then apply our evaluation functions to quantify characteristics such as edge fattening and surface smoothness. We demonstrate that our approach supports practitioners in selecting the most suitable algorithm for their application. Using the new Middlebury dataset, we show that rankings based on our metrics reveal specific algorithm strengths and weaknesses which are not quantified by existing metrics. We finally show how stacked bar charts and radar charts visually support multidimensional performance evaluation. An interactive stereo benchmark based on the proposed metrics and visualizations is available at: http://hci.iwr.uni-heidelberg.de/stereometrics.