Overcoming Domain Limitations in Open-vocabulary Segmentation

Dongjun Hwang, Seong Joon Oh, Junsuk Choe

Keywords: open-vocabulary, segmentation, fine-tuning, continual learning

TL;DR: We propose a novel decoder weight interpolation method to enhance the generalization capabilities of open-vocabulary segmentation models.

Abstract:

Open-vocabulary segmentation (OVS) has gained attention for its ability to recognize a broader range of classes. However, OVS models show significant performance drops when applied to target data distributions beyond the source dataset. Fine-tuning these models on new datasets can improve performance, but often leads to the catastrophic forgetting of previously learned knowledge. To address this issue, we propose a method that allows OVS models to learn information from new data distributions while preserving prior knowledge. Our approach begins by evaluating the input sample’s proximity to multiple data distributions, using precomputed multivariate normal distributions for each data distribution. Based on this prediction, we dynamically interpolate between the weights of the pre-trained decoder and the fine-tuned decoders. Extensive experiments demonstrate that this approach allows OVS models to adapt to new data distributions while maintaining performance on the source dataset.

Primary Area: transfer learning, meta learning, and lifelong learning

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 4976