Object-Proposal Evaluation Protocol is 'Gameable'

TL;DR

This paper critiques current object proposal evaluation methods for being 'gameable' and introduces a nearly-fully annotated dataset and diagnostic tools to better assess their true category independence and generalization.

Contribution

It presents a nearly-fully annotated dataset, evaluation protocols, and diagnostic tools to improve the assessment of object proposal algorithms beyond partial annotations.

Findings

Evaluation protocol is 'gameable' and does not reflect true category independence.

Nearly-fully annotated dataset reveals overfitting in existing methods.

Diagnostic tool detects bias capacity in object proposal algorithms.

Abstract

Object proposals have quickly become the de-facto pre-processing step in a number of vision pipelines (for object detection, object discovery, and other tasks). Their performance is usually evaluated on partially annotated datasets. In this paper, we argue that the choice of using a partially annotated dataset for evaluation of object proposals is problematic -- as we demonstrate via a thought experiment, the evaluation protocol is 'gameable', in the sense that progress under this protocol does not necessarily correspond to a "better" category independent object proposal algorithm. To alleviate this problem, we: (1) Introduce a nearly-fully annotated version of PASCAL VOC dataset, which serves as a test-bed to check if object proposal techniques are overfitting to a particular list of categories. (2) Perform an exhaustive evaluation of object proposal methods on our introduced nearly-fully annotated PASCAL dataset and perform cross-dataset generalization experiments; and (3) Introduce a diagnostic experiment to detect the bias capacity in an object proposal algorithm. This tool circumvents the need to collect a densely annotated dataset, which can be expensive and cumbersome to collect. Finally, we plan to release an easy-to-use toolbox which combines various publicly available implementations of object proposal algorithms which standardizes the proposal generation and evaluation so that new methods can be added and evaluated on different datasets. We hope that the results presented in the paper will motivate the community to test the category independence of various object proposal methods by carefully choosing the evaluation protocol.