Annotating Object Instances with a Polygon-RNN

TL;DR

This paper introduces a polygon prediction method for semi-automatic object annotation, significantly speeding up the process while maintaining high accuracy, and demonstrating good generalization across datasets.

Contribution

The paper presents a novel polygon-based approach for object annotation that allows human correction and achieves high accuracy and speed improvements over traditional pixel-labeling methods.

Findings

Speeds up annotation by a factor of 4.7 on Cityscapes.

Achieves 78.4% IoU agreement with ground-truth.

Generalizes well to unseen datasets.

Abstract

We propose an approach for semi-automatic annotation of object instances. While most current methods treat object segmentation as a pixel-labeling problem, we here cast it as a polygon prediction task, mimicking how most current datasets have been annotated. In particular, our approach takes as input an image crop and sequentially produces vertices of the polygon outlining the object. This allows a human annotator to interfere at any time and correct a vertex if needed, producing as accurate segmentation as desired by the annotator. We show that our approach speeds up the annotation process by a factor of 4.7 across all classes in Cityscapes, while achieving 78.4% agreement in IoU with original ground-truth, matching the typical agreement between human annotators. For cars, our speed-up factor is 7.3 for an agreement of 82.2%. We further show generalization capabilities of our approach to unseen datasets.