Publications

This work presents FOVEA, an add-on suite of analytic tools for the forensic review of video in large-scale surveillance systems. While significant investment has been made toward improving camera coverage and quality, the burden on video operators for reviewing and extracting useful information from the video has only increased. Daily investigation tasks (such as searching through video, investigating abandoned objects, or piecing together information from multiple cameras) still require a significant amount of manual review by video operators. In contrast to other tools which require exporting video data or otherwise curating the video collection before analysis, FOVEA is designed to integrate with existing surveillance systems. Tools can be applied to any video stream in an on-demand fashion without additional hardware. This paper details the technical approach, underlying algorithms, and effects on video operator performance.

Recent works in deep learning have been driven broadly by the desire to attain high accuracy on certain challenge problems. The network architecture and other hyperparameters of many published models are typically chosen by trial-and-error experiments with little considerations paid to resource constraints at deployment time. We propose a fully automated model learning approach that (1) treats architecture selection as part of the learning process, (2) uses a blend of broad-based random sampling and adaptive iterative refinement to explore the solution space, (3) performs optimization subject to given memory and computational constraints imposed by target deployment scenarios, and (4) is scalable and can use only a practically small number of GPUs for training. We present results that show graceful model degradation under strict resource constraints for object classification problems using CIFAR-10 in our experiments. We also discuss future work in further extending the approach.