Black-Box Optimization of Object Detector Scales
This work addresses the incremental improvement of object detection performance for computer vision researchers by automating hyper-parameter tuning, but it is incremental as it applies existing optimization methods to a new application.
The paper tackled the problem of manually tuned hyper-parameters in object detectors by applying black-box optimization methods to tune scales in Faster R-CNN and SSD, resulting in mAP increases of 2% on PASCAL VOC 2007 with Faster R-CNN and 3% with SSD, though with a 1% decrease in small objects on COCO.
Object detectors have improved considerably in the last years by using advanced CNN architectures. However, many detector hyper-parameters are generally manually tuned, or they are used with values set by the detector authors. Automatic Hyper-parameter optimization has not been explored in improving CNN-based object detectors hyper-parameters. In this work, we propose the use of Black-box optimization methods to tune the prior/default box scales in Faster R-CNN and SSD, using Bayesian Optimization, SMAC, and CMA-ES. We show that by tuning the input image size and prior box anchor scale on Faster R-CNN mAP increases by 2% on PASCAL VOC 2007, and by 3% with SSD. On the COCO dataset with SSD there are mAP improvement in the medium and large objects, but mAP decreases by 1% in small objects. We also perform a regression analysis to find the significant hyper-parameters to tune.