Segment Anything Model for automated image data annotation: empirical studies using text prompts from Grounding DINO

TL;DR

This paper investigates the use of Grounding DINO and SAM for automated image annotation across various domains, revealing predictable false positive patterns and demonstrating significant improvements in segmentation accuracy and efficiency.

Contribution

It provides empirical insights into false positive patterns in REC-based detection and demonstrates how size-based filtering improves segmentation in specialized domains.

Findings

False positives tend to be large and can be filtered by size.

Size-based filtering reduces false positives and improves segmentation accuracy.

SAM significantly enhances annotation efficiency and accuracy.

Abstract

Grounding DINO and the Segment Anything Model (SAM) have achieved impressive performance in zero-shot object detection and image segmentation, respectively. Together, they have a great potential to revolutionize applications in zero-shot semantic segmentation or data annotation. Yet, in specialized domains like medical image segmentation, objects of interest (e.g., organs, tissues, and tumors) may not fall in existing class names. To address this problem, the referring expression comprehension (REC) ability of Grounding DINO is leveraged to detect arbitrary targets by their language descriptions. However, recent studies have highlighted severe limitation of the REC framework in this application setting owing to its tendency to make false positive predictions when the target is absent in the given image. And, while this bottleneck is central to the prospect of open-set semantic segmentation, it is still largely unknown how much improvement can be achieved by studying the prediction errors. To this end, we perform empirical studies on six publicly available datasets across different domains and reveal that these errors consistently follow a predictable pattern and can, thus, be mitigated by a simple strategy. Specifically, we show that false positive detections with appreciable confidence scores generally occupy large image areas and can usually be filtered by their relative sizes. More importantly, we expect these observations to inspire future research in improving REC-based detection and automated segmentation. Meanwhile, we evaluate the performance of SAM on multiple datasets from various specialized domains and report significant improvements in segmentation performance and annotation time savings over manual approaches.