Open Vocabulary Semantic Segmentation with Patch Aligned Contrastive Learning

By Jishnu Mukhoti, Tsung-Yu Lin, Omid Poursaeed, Rui Wang, Ashish Shah, Philip H. S. Torr, Ser-Nam Lim. 2022

144 citations recorded by the source corpus.

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.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jianfeng Gao
2. Jianwei Yang
3. Lei Zhang
4. Xiaohua Zhai
5. Basil Mustafa
6. Lucas Beyer
7. Neil Houlsby
8. Alexander Kolesnikov

In Computer Vision and Pattern Recognition

1. Luc Van Gool
2. Dacheng Tao
3. Lei Zhang
4. Trevor Darrell
5. Ming-Hsuan Yang
6. Jianfeng Gao
7. Xiaogang Wang
8. Jitendra Malik

In Radiology, Nuclear Medicine and Imaging

1. Lucas Beyer
2. Jianwei Yang
3. Ken Goldberg
4. Alexander Kolesnikov
5. Jianfeng Gao
6. Angjoo Kanazawa
7. Feng Li
8. Hao Zhang

In Representation & Self-Supervised Learning

1. Xiaohua Zhai
2. Lucas Beyer
3. Alexander Kolesnikov
4. Basil Mustafa
5. Lei Zhang
6. Jianfeng Gao
7. Jianwei Yang
8. Ken Goldberg

Citations in the corpus

• A Simple Framework for Open-Vocabulary Segmentation and Detection (2023, background): “…large amount of image-text data for open-vocabulary detection or segmentation, such as distilling the visual-semantic representations from multi-modal foundation models [12, 57], designing fine-grained or augmented contrastive learning methods [39] or utilizing pseudo-labeling techniques [30, 56].”
• Sigmoid Loss for Language Image Pre-Training (2023, background): “Follow-up works show that contrastively pre-trained models produce good representations for fine-tuning [39, 12], linear regression [18], object detection [25], semantic segmentation [26] and video tasks [43].”
• LERF: Language Embedded Radiance Fields (2023, background): “Con-Volume sequently, as noted by the authors of OV-Seg [22], such networks often face difficulties in accurately segmenting unlabeled hierarchical components of the original masks, such as object parts.”