Search

Object detection with multimodal inputs can improve many safety-critical systems such as autonomous vehicles (AVs). Motivated by AVs that operate in both day and night, we study multimodal object detection with RGB and thermal cameras, since the latter provides much stronger object signatures under poor illumination. We explore strategies for fusing information from different modalities. Our key contribution is a probabilistic ensembling technique, ProbEn, a simple non-learned method that fuses together detections from multi-modalities. We derive ProbEn from Bayes' rule and first principles that assume conditional independence across modalities. Through probabilistic marginalization, ProbEn elegantly handles missing modalities when detectors do not fire on the same object. Importantly, ProbEn also notably improves multimodal detection even when the conditional independence assumption does not hold, e.g., fusing outputs from other fusion methods (both off-the-shelf and trained in-house). We validate ProbEn on two benchmarks containing both aligned (KAIST) and unaligned (FLIR) multimodal images, showing that ProbEn outperforms prior work by more than 13% in relative performance!, Comment: camera-ready with supplement for ECCV2022 (oral presentation); open-source code at https://github.com/Jamie725/RGBT-detection

Over the last four years we have operated a public smartphone platform testbed called PHONELAB. PHONELAB consists of up to several-hundred participants who run an experimental platform image on their primary smartphone. The experimental platform consists of both instrumentation and experimental changes to platform components, including core Android services and Linux. This paper describes the design of the testbed, the process of conducting PHONELAB experiments, and some of the research the testbed has supported. We also offer many lessons learned along the way, almost all of which have been learned the hard way--through trial and a lot of error. We expect our experiences will help those contemplating operating large user-facing testbeds, anyone conducting experiments on smartphones, and many mobile systems researchers.

Validating wireless protocol implementations is challenging. Today's approaches require labor-intensive experimental setup and manual trace investigation, but produce poor coverage and inaccurate and irreproducible results. We present VERIFI, the first systematic sniffer-based, model-guided runtime verification framework for wireless protocol implementations. VERIFI takes a formal model of the protocol being verified as input. To achieve good coverage, it first applies state reachability analysis by applying model checking techniques. It then uses a new PACKETSNIPER component to selectively trigger packet losses required to quickly investigate all reachable protocol states. Our results show that the selective packet jamming allows VERIFI to significantly improve the coverage of protocol states. Finally, VERIFI accommodates uncertainty caused by the sniffer when validating traces, allowing it to provide accurate and reproducible results. By modeling uncertainty, VERIFI highlights likely protocol violations for developers to examine.