Make VLM Recognize Visual Hallucination on Cartoon Character Image with Pose Information

2024-03-22

AI Summary

Large-scale text-to-image models frequently generate structural visual hallucinations—such as limb misalignment and anatomical disproportion—in cartoon/pixel-art images, yet these artifacts remain difficult to detect reliably. To address this, we propose Pose-Aware In-Context Visual Learning (PA-ICVL), the first method to integrate human pose estimation into the in-context learning framework of vision-language models (VLMs). PA-ICVL jointly encodes RGB and pose features, enabling fine-grained identification of semantic-structural hallucinations in non-photorealistic images with only few-shot examples. Evaluated on our newly constructed Cartoon Hallucination Synthesis Dataset, PA-ICVL achieves hallucination detection accuracies of 78% and 80% for GPT-4V and Gemini Pro Vision, respectively—improving over RGB-only baselines by 50–57%. We publicly release both the dataset and a lightweight fine-tuned model, facilitating user-customized VLM specialization for non-photorealistic image understanding.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Leveraging large-scale Text-to-Image (TTI) models have become a common technique for generating exemplar or training dataset in the fields of image synthesis, video editing, 3D reconstruction. However, semantic structural visual hallucinations involving perceptually severe defects remain a concern, especially in the domain of non-photorealistic rendering (NPR) such as cartoons and pixelization-style character. To detect these hallucinations in NPR, We propose a novel semantic structural hallucination detection system using Vision-Language Model (VLM). Our approach is to leverage the emerging capability of large language model, in-context learning which denotes that VLM has seen some examples by user for specific downstream task, here hallucination detection. Based on in-context learning, we introduce pose-aware in-context visual learning (PA-ICVL) which improve the overall performance of VLM by further inputting visual data beyond prompts, RGB images and pose information. By incorporating pose guidance, we enable VLMs to make more accurate decisions. Experimental results demonstrate significant improvements in identifying visual hallucinations compared to baseline methods relying solely on RGB images. Within selected two VLMs, GPT-4v, Gemini pro vision, our proposed PA-ICVL improves the hallucination detection with 50% to 78%, 57% to 80%, respectively. This research advances a capability of TTI models toward real-world applications by mitigating visual hallucinations via in-context visual learning, expanding their potential in non-photorealistic domains. In addition, it showcase how users can boost the downstream-specialized capability of open VLM by harnessing additional conditions. We collect synthetic cartoon-hallucination dataset with TTI models, this dataset and final tuned VLM will be publicly available.

Problem

• Non-photorealistic Rendering
• Text-to-Image Synthesis
• Visual Artifact

Innovation

• pose-aware contextual visual learning
• visual illusion detection
• non-realistic imagery