Abstract: Deep video action recognition models have been highly successful in recent years but require large quantities of manually-annotated data, which are expensive and laborious to obtain. In this work, we investigate the generation of synthetic training data for video action recognition, as synthetic data have been successfully used to supervise models for a variety of other computer vision tasks. We propose an interpretable parametric generative model of human action videos that relies on procedural generation, physics models and other components of modern game engines. With this model we generate a diverse, realistic, and physically plausible dataset of human action videos, called PHAV for “Procedural Human Action Videos”. PHAV contains a total of 39,982 videos, with more than 1000 examples for each of 35 action categories. Our video generation approach is not limited to existing motion capture sequences: 14 of these 35 categories are procedurally-defined synthetic actions. In addition, each video is represented with 6 different data modalities, including RGB, optical flow and pixel-level semantic labels. These modalities are generated almost simultaneously using the Multiple Render Targets feature of modern GPUs. In order to leverage PHAV, we introduce a deep multi-task (i.e. that considers action classes from multiple datasets) representation learning architecture that is able to simultaneously learn from synthetic and real video datasets, even when their action categories differ. Our experiments on the UCF-101 and HMDB-51 benchmarks suggest that combining our large set of synthetic videos with small real-world datasets can boost recognition performance. Our approach also significantly outperforms video representations produced by fine-tuning state-of-the-art unsupervised generative models of videos.

Abstract: Detection of lane markings based on a camera sensor can be a low-cost solution to lane departure and curve-over-speed warnings. A number of methods and implementations have been reported in the literature. However, reliable detection is still an issue because of cast shadows, worn and occluded markings, variable ambient lighting conditions, for example. We focus on increasing detection reliability in two ways. First, we employed an image feature other than the commonly used edges: ridges, which we claim addresses this problem better. Second, we adapted RANSAC, a generic robust estimation method, to fit a parametric model of a pair of lane lines to the image features, based on both ridgeness and ridge orientation. In addition, the model was fitted for the left and right lane lines simultaneously to enforce a consistent result. Four measures of interest for driver assistance applications were directly computed from the fitted parametric model at each frame: lane width, lane curvature, and vehicle yaw angle and lateral offset with regard the lane medial axis. We qualitatively assessed our method in video sequences captured on several road types and under very different lighting conditions. We also quantitatively assessed it on synthetic but realistic video sequences for which road geometry and vehicle trajectory ground truth are known.