Less Is More: Linear Layers on CLIP Features as Powerful VizWiz Model

TL;DR

This paper introduces a simplified CLIP-based model for visual question answering that uses linear layers without fine-tuning, achieving competitive results with reduced complexity and computational requirements.

Contribution

The authors propose a minimalistic architecture using linear classifiers on CLIP features, eliminating the need for fine-tuning and incorporating an auxiliary loss for improved performance.

Findings

Achieved 60.15% accuracy on VizWiz Task 1

Attained 83.78% AP score on VizWiz Task 2

Reduced model complexity and training requirements

Abstract

Current architectures for multi-modality tasks such as visual question answering suffer from their high complexity. As a result, these architectures are difficult to train and require high computational resources. To address these problems we present a CLIP-based architecture that does not require any fine-tuning of the feature extractors. A simple linear classifier is used on the concatenated features of the image and text encoder. During training an auxiliary loss is added which operates on the answer types. The resulting classification is then used as an attention gate on the answer class selection. On the VizWiz 2022 Visual Question Answering Challenge we achieve 60.15 % accuracy on Task 1: Predict Answer to a Visual Question and AP score of 83.78 % on Task 2: Predict Answerability of a Visual Question.