Thinking Like an Annotator: Generation of Dataset Labeling Instructions

TL;DR

This paper introduces a new task called Labeling Instruction Generation to create detailed dataset labeling instructions, using a no-training retrieval framework that improves annotation quality and transparency for large-scale datasets.

Contribution

It proposes a novel framework leveraging pre-trained vision-language models to generate and evaluate dataset labeling instructions without model training.

Findings

Framework outperforms baselines with 7.06 mAP improvement on NuImages

Framework achieves 12.9 mAP increase on COCO

Generates diverse visual and text representations of dataset categories

Abstract

Large-scale datasets are essential to modern day deep learning. Advocates argue that understanding these methods requires dataset transparency (e.g. "dataset curation, motivation, composition, collection process, etc..."). However, almost no one has suggested the release of the detailed definitions and visual category examples provided to annotators - information critical to understanding the structure of the annotations present in each dataset. These labels are at the heart of public datasets, yet few datasets include the instructions that were used to generate them. We introduce a new task, Labeling Instruction Generation, to address missing publicly available labeling instructions. In Labeling Instruction Generation, we take a reasonably annotated dataset and: 1) generate a set of examples that are visually representative of each category in the dataset; 2) provide a text label that corresponds to each of the examples. We introduce a framework that requires no model training to solve this task and includes a newly created rapid retrieval system that leverages a large, pre-trained vision and language model. This framework acts as a proxy to human annotators that can help to both generate a final labeling instruction set and evaluate its quality. Our framework generates multiple diverse visual and text representations of dataset categories. The optimized instruction set outperforms our strongest baseline across 5 folds by 7.06 mAP for NuImages and 12.9 mAP for COCO.