AdaScale: Towards Real-time Video Object Detection Using Adaptive Scaling
This addresses the need for real-time, accurate object detection in autonomous systems like robots and cars, offering a novel approach that avoids the typical speed-accuracy trade-off.
The paper tackles video object detection by proposing AdaScale, which adaptively selects input image scales to improve both accuracy and speed, achieving up to 2.7 points mAP improvement with 1.8x speedup on datasets like ImageNet VID.
In vision-enabled autonomous systems such as robots and autonomous cars, video object detection plays a crucial role, and both its speed and accuracy are important factors to provide reliable operation. The key insight we show in this paper is that speed and accuracy are not necessarily a trade-off when it comes to image scaling. Our results show that re-scaling the image to a lower resolution will sometimes produce better accuracy. Based on this observation, we propose a novel approach, dubbed AdaScale, which adaptively selects the input image scale that improves both accuracy and speed for video object detection. To this end, our results on ImageNet VID and mini YouTube-BoundingBoxes datasets demonstrate 1.3 points and 2.7 points mAP improvement with 1.6x and 1.8x speedup, respectively. Additionally, we improve state-of-the-art video acceleration work by an extra 1.25x speedup with slightly better mAP on ImageNet VID dataset.