Open Vocabulary Semantic Segmentation with Patch Aligned Contrastive Learning

2022/12/09 by Jishnu Mukhoti, Tsung‐Yu Lin, Mukhoti, Jishnu +11 · 20 citations
Computer Science · Medicine · #COVID-19 diagnosis using AI #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Multimodal Machine Learning Applications

paper · pdf · doi:10.48550/arxiv.2212.04994

openalex publication_date 2022/12/09 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We introduce Patch Aligned Contrastive Learning (PACL), a modified compatibility function for CLIP's contrastive loss, intending to train an alignment between the patch tokens of the vision encoder and the CLS token of the text encoder. With such an alignment, a model can identify regions of an image corresponding to a given text input, and therefore transfer seamlessly to the task of open vocabulary semantic segmentation without requiring any segmentation annotations during training. Using pre-trained CLIP encoders with PACL, we are able to set the state-of-the-art on the task of open vocabulary zero-shot segmentation on 4 different segmentation benchmarks: Pascal VOC, Pascal Context, COCO Stuff and ADE20K. Furthermore, we show that PACL is also applicable to image-level predictions and when used with a CLIP backbone, provides a general improvement in zero-shot classification accuracy compared to CLIP, across a suite of 12 image classification datasets.

Cited by

• Retrieve and Segment: Are a Few Examples Enough to Bridge the Supervision Gap in Open-Vocabulary Segmentation?
• β-CLIP: Text-Conditioned Contrastive Learning for Multi-Granular Vision-Language Alignment
• PowerCLIP: Powerset Alignment for Contrastive Pre-Training
• FarSLIP: Discovering Effective CLIP Adaptation for Fine-Grained Remote Sensing Understanding
• Improving Visual Discriminability of CLIP for Training-Free Open-Vocabulary Semantic Segmentation
• One Patch to Caption Them All: A Unified Zero-Shot Captioning Framework
• AttAnchor: Guiding Cross-Modal Token Alignment in VLMs with Attention Anchors
• Enhancing Vehicle Detection under Adverse Weather Conditions with Contrastive Learning
• Cross-modal Full-mode Fine-grained Alignment for Text-to-Image Person Retrieval
• Synthetic Captions for Open-Vocabulary Zero-Shot Segmentation
• Multi-View Slot Attention Using Paraphrased Texts for Face Anti-Spoofing
• Plug-in Feedback Self-adaptive Attention in CLIP for Training-free Open-Vocabulary Segmentation
• Annotation-Free Open-Vocabulary Segmentation for Remote-Sensing Images
• Generalized Decoupled Learning for Enhancing Open-Vocabulary Dense Perception
• VesselRW: Weakly Supervised Subcutaneous Vessel Segmentation via Learned Random Walk Propagation
• DualResolution Residual Architecture with Artifact Suppression for Melanocytic Lesion Segmentation
• Edge Detection for Organ Boundaries via Top Down Refinement and SubPixel Upsampling
• SynSeg: Feature Synergy for Multi-Category Contrastive Learning in End-to-End Open-Vocabulary Semantic Segmentation
• What Holds Back Open-Vocabulary Segmentation?
• Deeply Dual Supervised learning for melanoma recognition

Related

• Explore the Potential of CLIP for Training-Free Open Vocabulary Semantic Segmentation
• CorrCLIP: Reconstructing Patch Correlations in CLIP for Open-Vocabulary Semantic Segmentation
• SegCLIP: Patch Aggregation with Learnable Centers for Open-Vocabulary Semantic Segmentation
• TagCLIP: Improving Discrimination Ability of Open-Vocabulary Semantic Segmentation
• Harnessing Vision Foundation Models for High-Performance, Training-Free Open Vocabulary Segmentation