Aligning Object Detector Bounding Boxes with Human Preference

TL;DR

This paper investigates how to align object detector bounding boxes with human preferences, proposing an asymmetric loss to improve the perceptual quality of detections, validated through user studies and qualitative analysis.

Contribution

It introduces an asymmetric bounding box regression loss that better aligns detector outputs with human preferences, improving perceptual quality.

Findings

Humans prefer upscaled bounding boxes by factors of 1.5 or 2.

Detectors with the asymmetric loss produce boxes more aligned with human preferences.

Human preference may be influenced by object shape characteristics.

Abstract

Previous work shows that humans tend to prefer large bounding boxes over small bounding boxes with the same IoU. However, we show here that commonly used object detectors predict large and small boxes equally often. In this work, we investigate how to align automatically detected object boxes with human preference and study whether this improves human quality perception. We evaluate the performance of three commonly used object detectors through a user study (N = 123). We find that humans prefer object detections that are upscaled with factors of 1.5 or 2, even if the corresponding AP is close to 0. Motivated by this result, we propose an asymmetric bounding box regression loss that encourages large over small predicted bounding boxes. Our evaluation study shows that object detectors fine-tuned with the asymmetric loss are better aligned with human preference and are preferred over fixed scaling factors. A qualitative evaluation shows that human preference might be influenced by some object characteristics, like object shape.