Vision language models speed up error detection in image classifiers
Beyond Benchmarks: Using VLMs to Reveal Systematic Classification Failures Under Real World Conditions
Computer Vision and Pattern RecognitionArtificial IntelligenceEmerging Technologies
Summary
Checking where image recognition systems make mistakes usually involves a lot of manual work. The authors looked at using vision language models, which understand both images and words, to group and describe these errors automatically. They found these models can spot added errors in regular images and help cluster images in defense-related ones, though with some overlap. Their work shows this approach could make error checking faster but isn’t reliable enough yet to replace humans.
What this means in practice
- •For sensor data engineers: Group and describe recurring classification errors automatically to speed up error analysis in sensor data systems.
- •For security system developers: Use vision language models to help cluster and explain errors in military image classifiers, improving monitoring during deployment.
Tested on one dataset.
Authors
Dieuwertje Alblas, Alma M. Liezenga, Jan Erik van Woerden, Fedor Taggenbrock, Dalia Aljawaheri, Klamer Schutte
Abstract
Verification and validation (V&V) of classification models is crucial to enable a wide range of sensor processing applications. Currently, the V&V process relies on time-consuming manual inspection of erroneous samples to find meaningful patterns. This work explores the use of Vision Language Models (VLMs) to speed up this laborious process. VLMs are trained to embed images into a semantically meaningful vector representation, from which human-interpretable systematic errors can be distilled. Deploying such VLM-based methods in a defence context introduces two major challenges: (1) the defence domain is underrepresented in the training data of VLMs, and (2) surroundings and context are less diverse than for other domains. This study provides an initial assessment of the suitability of VLM-based methods for V&V of defence applications. We propose a VLM-based error slice detection (ESD) method that independently groups and labels systematic errors made by a classification model. We demonstrate that this method is able to identify operationally-relevant artificially added perturbations in a non-military dataset. In a military context, our method clusters and describes images based on their surroundings, but also exhibits overlap between cluster descriptions. We further investigate the difference in embedding variation between our military and non-military dataset, which remains a topic of interest. Although the results do not yet warrant fully automated V&V through VLM-based ESD, they show that VLMs could be used to accelerate V&V processes in the future.