Abstract-This paper reports on WaterGAN, a generative adversarial network (GAN) for generating realistic underwater images from in-air image and depth pairings in an unsupervised pipeline used for color correction of monocular underwater images. Cameras onboard autonomous and remotely operated vehicles can capture high resolution images to map the seafloor; however, underwater image formation is subject to the complex process of light propagation through the water column. The raw images retrieved are characteristically different than images taken in air due to effects such as absorption and scattering, which cause attenuation of light at different rates for different wavelengths. While this physical process is well described theoretically, the model depends on many parameters intrinsic to the water column as well as the structure of the scene. These factors make recovery of these parameters difficult without simplifying assumptions or field calibration; hence, restoration of underwater images is a non-trivial problem. Deep learning has demonstrated great success in modeling complex nonlinear systems but requires a large amount of training data, which is difficult to compile in deep sea environments. Using WaterGAN, we generate a large training dataset of corresponding depth, in-air color images, and realistic underwater images. This data serves as input to a two-stage network for color correction of monocular underwater images. Our proposed pipeline is validated with testing on real data collected from both a pure water test tank and from underwater surveys collected in the field. Source code, sample datasets, and pretrained models are made publicly available.

Recent work has focused on generating synthetic imagery to increase the size and variability of training data for learning visual tasks in urban scenes. This includes increasing the occurrence of occlusions or varying environmental and weather effects. However, few have addressed modeling variation in the sensor domain. Sensor effects can degrade real images, limiting generalizability of network performance on visual tasks trained on synthetic data and tested in real environments. This paper proposes an efficient, automatic, physically-based augmentation pipeline to vary sensor effects -chromatic aberration, blur, exposure, noise, and color temperature -for synthetic imagery. In particular, this paper illustrates that augmenting synthetic training datasets with the proposed pipeline reduces the domain gap between synthetic and real domains for the task of object detection in urban driving scenes.

Performance on benchmark datasets has drastically improved with advances in deep learning. Still, crossdataset generalization performance remains relatively low due to the domain shift that can occur between two different datasets. This domain shift is especially exaggerated between synthetic and real datasets. Significant research has been done to reduce this gap, specifically via modeling variation in the spatial layout of a scene, such as occlusions, and scene environmental factors, such as time of day and weather effects. However, few works have addressed modeling the variation in the sensor domain as a means of reducing the synthetic to real domain gap. The camera or sensor used to capture a dataset introduces artifacts into the image data that are unique to the sensor model, suggesting that sensor effects may also contribute to domain shift. To address this, we propose a learned augmentation network composed of physically-based augmentation functions. Our proposed augmentation pipeline transfers specific effects of the sensor model -chromatic aberration, blur, exposure, noise, and color temperature -from a real dataset to a synthetic dataset. We provide experiments that demonstrate that augmenting synthetic training datasets with the proposed learned augmentation framework reduces the domain gap between synthetic and real domains for object detection in urban driving scenes. 1 A. Carlson and K. A. Skinner are with the Robotics Institute, University of Michigan,

Recent work has shown that convolutional neural networks (CNNs) can be applied successfully in disparity estimation, but these methods still suffer from errors in regions of low-texture, occlusions and reflections. Concurrently, deep learning for semantic segmentation has shown great progress in recent years. In this paper, we design a CNN architecture that combines these two tasks to improve the quality and accuracy of disparity estimation with the help of semantic segmentation. Specifically, we propose a network structure in which these two tasks are highly coupled. One key novelty of this approach is the two-stage refinement process. Initial disparity estimates are refined with an embedding learned from the semantic segmentation branch of the network. The proposed model is trained using an unsupervised approach, in which images from one half of the stereo pair are warped and compared against images from the other camera. Another key advantage of the proposed approach is that a single network is capable of outputting disparity estimates and semantic labels. These outputs are of great use in autonomous vehicle operation; with real-time constraints being key, such performance improvements increase the viability of driving applications. Experiments on KITTI and Cityscapes datasets show that our model can achieve state-ofthe-art results and that leveraging embedding learned from semantic segmentation improves the performance of disparity estimation.

Studies of behavioral momentum reveal that reinforcing an alternative response in the presence of a target response reduces the rate of target responding but increases its persistence, relative to training the target response on its own. Because of the parallels between these studies and differential-reinforcement techniques to reduce problem behavior in clinical settings, alternative techniques to reduce problem behavior without enhancing its persistence are being explored. One potential solution is to train an alternative response in a separate stimulus context from problem behavior before combining the alternative stimulus with the target stimulus. The present study assessed how differences in reinforcement contingencies and rate for alternative responding influenced resistance to extinction of target responding when combining alternative and target stimuli in pigeons. Across three experiments, alternative stimuli signaling a response-reinforcer dependency and greater reinforcer rates more effectively decreased the persistence of target responding when combining alternative and target stimuli within the same extinction tests, but not when compared across separate extinction tests. Overall, these findings reveal that differences in competition between alternative and target responding produced by contingencies of alternative reinforcement could influence the effectiveness of treating problem behavior through combining stimulus contexts.