If a Human Can See It, So Should Your System: Reliability Requirements for Machine Vision Components

TL;DR

This paper proposes a human-inspired, machine-verifiable framework for defining and testing reliability requirements of machine vision components, ensuring safety-critical robustness against environmental transformations.

Contribution

It introduces a novel approach to specify and verify reliability requirements for MVCs based on human performance benchmarks and safety-related image transformations.

Findings

Reliability requirements can be instantiated from human performance data.

The method detects reliability gaps in MVCs that other methods miss.

Feasibility demonstrated on 13 state-of-the-art models.

Abstract

Machine Vision Components (MVC) are becoming safety-critical. Assuring their quality, including safety, is essential for their successful deployment. Assurance relies on the availability of precisely specified and, ideally, machine-verifiable requirements. MVCs with state-of-the-art performance rely on machine learning (ML) and training data but largely lack such requirements. In this paper, we address the need for defining machine-verifiable reliability requirements for MVCs against transformations that simulate the full range of realistic and safety-critical changes in the environment. Using human performance as a baseline, we define reliability requirements as: 'if the changes in an image do not affect a human's decision, neither should they affect the MVC's.' To this end, we provide: (1) a class of safety-related image transformations; (2) reliability requirement classes to specify correctness-preservation and prediction-preservation for MVCs; (3) a method to instantiate machine-verifiable requirements from these requirements classes using human performance experiment data; (4) human performance experiment data for image recognition involving eight commonly used transformations, from about 2000 human participants; and (5) a method for automatically checking whether an MVC satisfies our requirements. Further, we show that our reliability requirements are feasible and reusable by evaluating our methods on 13 state-of-the-art pre-trained image classification models. Finally, we demonstrate that our approach detects reliability gaps in MVCs that other existing methods are unable to detect.