Test-time Contrastive Concepts for Open-world Semantic Segmentation with Vision-Language Models
This addresses a more realistic scenario for open-world segmentation, but is incremental as it builds on existing CLIP-like models and focuses on improving contrastive concept generation.
The paper tackles the problem of open-world semantic segmentation with vision-language models when only a single textual prompt is provided, by proposing methods to automatically generate query-specific contrastive concepts at test time using the VLM's training distribution or LLM prompts, and shows relevance on common datasets with a new evaluation metric.
Recent CLIP-like Vision-Language Models (VLMs), pre-trained on large amounts of image-text pairs to align both modalities with a simple contrastive objective, have paved the way to open-vocabulary semantic segmentation. Given an arbitrary set of textual queries, image pixels are assigned the closest query in feature space. However, this works well when a user exhaustively lists all possible visual concepts in an image that contrast against each other for the assignment. This corresponds to the current evaluation setup in the literature, which relies on having access to a list of in-domain relevant concepts, typically classes of a benchmark dataset. Here, we consider the more challenging (and realistic) scenario of segmenting a single concept, given a textual prompt and nothing else. To achieve good results, besides contrasting with the generic 'background' text, we propose two different approaches to automatically generate, at test time, query-specific textual contrastive concepts. We do so by leveraging the distribution of text in the VLM's training set or crafted LLM prompts. We also propose a metric designed to evaluate this scenario and show the relevance of our approach on commonly used datasets.