Vehicle detection in aerial imagery is a challenging task due to small object sizes, high object density and partial occlusions. While past research mostly focused on improving detection accuracy, inference speed is another important factor when using CNN object detectors in a real life scenario -especially when targeting mobile platforms like unmanned aerial vehicles (UAVs). In this work, we compare several established detection frameworks in terms of their accuracy-speed trade-off and show that the Single Shot MultiBox Detector (SSD) offers the best compromise. We subsequently undertake a thorough evaluation of several design choices to further increase detection speed while sacrificing little to no accuracy. This includes the choice of base network architecture, improved prediction layers and an automatic model pruning approach. Given our evaluation results, we finally construct UAV-Net -a novel aerial vehicle detector that has a model size of less than 0.4 MiB and is more than 16 times faster than current top performing approaches. UAV-Net is well suited for on-board processing and operates in real time on a Jetson TX2 platform. Nevertheless, its accuracy is on par with state-of-the-art approaches on the DLR 3K, VEDAI and UAVDT datasets. Code and models are available on the project website. 1
Unsupervised domain adaptation (UDA) deals with the adaptation process of a model to an unlabeled target domain while annotated data is only available for a given source domain. This poses a challenging task, as the domain shift between source and target instances deteriorates a model's performance when not addressed. In this paper, we propose UBR 2 S -the Uncertainty-Based Resampling and Reweighting Strategy -to tackle this problem. UBR 2 S employs a Monte Carlo dropout-based uncertainty estimate to obtain per-class probability distributions, which are then used for dynamic resampling of pseudo-labels and reweighting based on their sample likelihood and the accompanying decision error. Our proposed method achieves state-of-the-art results on multiple UDA datasets with single and multi-source adaptation tasks and can be applied to any off-theshelf network architecture. Code for our method is available at https://gitlab. com/tringwald/UBR2S.
Unsupervised domain adaptation (UDA) deals with the problem of classifying unlabeled target domain data while labeled data is only available for a different source domain. Unfortunately, commonly used classification methods cannot fulfill this task adequately due to the domain gap between the source and target data. In this paper, we propose a novel uncertainty-aware domain adaptation setup that models uncertainty as a multivariate Gaussian distribution in feature space. We show that our proposed uncertainty measure correlates with other common uncertainty quantifications and relates to smoothing the classifier's decision boundary, therefore improving the generalization capabilities. We evaluate our proposed pipeline on challenging UDA datasets and achieve state-of-the-art results. Code for our method is available at https://gitlab.com/ tringwald/cvp.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.