BEA: Revisiting anchor-based object detection DNN using Budding Ensemble Architecture

TL;DR

This paper presents BEA, a novel ensemble architecture for anchor-based object detection that improves confidence calibration, reduces uncertainty errors, and enhances detection accuracy and out-of-distribution detection across multiple datasets.

Contribution

BEA introduces a reduced ensemble architecture with new loss functions that improve confidence calibration and detection accuracy in anchor-based models.

Findings

BEA improves mAP and AP50 on KITTI dataset by 6% and 3.7%.

BEA achieves a 9.6% higher AP50 with better uncertainty thresholding.

BEA enhances out-of-distribution detection on multiple datasets.

Abstract

This paper introduces the Budding Ensemble Architecture (BEA), a novel reduced ensemble architecture for anchor-based object detection models. Object detection models are crucial in vision-based tasks, particularly in autonomous systems. They should provide precise bounding box detections while also calibrating their predicted confidence scores, leading to higher-quality uncertainty estimates. However, current models may make erroneous decisions due to false positives receiving high scores or true positives being discarded due to low scores. BEA aims to address these issues. The proposed loss functions in BEA improve the confidence score calibration and lower the uncertainty error, which results in a better distinction of true and false positives and, eventually, higher accuracy of the object detection models. Both Base-YOLOv3 and SSD models were enhanced using the BEA method and its proposed loss functions. The BEA on Base-YOLOv3 trained on the KITTI dataset results in a 6% and 3.7% increase in mAP and AP50, respectively. Utilizing a well-balanced uncertainty estimation threshold to discard samples in real-time even leads to a 9.6% higher AP50 than its base model. This is attributed to a 40% increase in the area under the AP50-based retention curve used to measure the quality of calibration of confidence scores. Furthermore, BEA-YOLOV3 trained on KITTI provides superior out-of-distribution detection on Citypersons, BDD100K, and COCO datasets compared to the ensembles and vanilla models of YOLOv3 and Gaussian-YOLOv3.