ROMA: Run-Time Object Detection To Maximize Real-Time Accuracy
This work addresses real-time accuracy optimization for object detection in dynamic video scenarios, representing an incremental improvement over existing runtime techniques.
The paper tackles the problem of real-time object detection accuracy degradation due to varying video content and detection latency by proposing ROMA, a run-time model that dynamically selects the optimal detector from a set, achieving accuracy improvements of 4 to 37% on MOT17Det and MOT20Dat datasets compared to baseline methods.
This paper analyzes the effects of dynamically varying video contents and detection latency on the real-time detection accuracy of a detector and proposes a new run-time accuracy variation model, ROMA, based on the findings from the analysis. ROMA is designed to select an optimal detector out of a set of detectors in real time without label information to maximize real-time object detection accuracy. ROMA utilizing four YOLOv4 detectors on an NVIDIA Jetson Nano shows real-time accuracy improvements by 4 to 37% for a scenario of dynamically varying video contents and detection latency consisting of MOT17Det and MOT20Det datasets, compared to individual YOLOv4 detectors and two state-of-the-art runtime techniques.