# Contrastive pretraining for semantic segmentation is robust to noisy positive pairs

arXiv: https://arxiv.org/abs/2211.13756 · HTML: https://mlsanity.com/paper/2211.13756/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 33 references that are in the map.
- First posted on arXiv 24 Nov 2022.

## Builds on (most cited first, 24 of 24)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [A Simple Framework for Contrastive Learning of Visual Representations](https://mlsanity.com/paper/2002.05709/) (2020)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Momentum Contrast for Unsupervised Visual Representation Learning](https://mlsanity.com/paper/1911.05722/) (2019)
- [Fully Convolutional Networks for Semantic Segmentation](https://mlsanity.com/paper/1411.4038/) (2014)
- [Bootstrap your own latent: A new approach to self-supervised Learning](https://mlsanity.com/paper/2006.07733/) (2020)
- [CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Features](https://mlsanity.com/paper/1905.04899/) (2019)
- [Improved Baselines with Momentum Contrastive Learning](https://mlsanity.com/paper/2003.04297/) (2020)
- [Describing Textures in the Wild](https://mlsanity.com/paper/1311.3618/) (2013)
- [Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination](https://mlsanity.com/paper/1805.01978/) (2018)
- [Dense Contrastive Learning for Self-Supervised Visual Pre-Training](https://mlsanity.com/paper/2011.09157/) (2020)
- [Contrastive learning of global and local features for medical image segmentation with limited annotations](https://mlsanity.com/paper/2006.10511/) (2020)
- [Unsupervised Semantic Segmentation by Distilling Feature Correspondences](https://mlsanity.com/paper/2203.08414/) (2022)
- [Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data](https://mlsanity.com/paper/2103.16607/) (2021)
- [Geography-Aware Self-Supervised Learning](https://mlsanity.com/paper/2011.09980/) (2020)
- [Boosting Contrastive Self-Supervised Learning with False Negative Cancellation](https://mlsanity.com/paper/2011.11765/) (2020)
- [Pixel Contrastive-Consistent Semi-Supervised Semantic Segmentation](https://mlsanity.com/paper/2108.09025/) (2021)
- [Contrastive Learning for Label-Efficient Semantic Segmentation](https://mlsanity.com/paper/2012.06985/) (2020)
- [SSL4EO-S12: A Large-Scale Multi-Modal, Multi-Temporal Dataset for Self-Supervised Learning in Earth Observation](https://mlsanity.com/paper/2211.07044/) (2022)
- [A Simple Data Mixing Prior for Improving Self-Supervised Learning](https://mlsanity.com/paper/2206.07692/) (2022)
- [Self-supervised pre-training enhances change detection in Sentinel-2 imagery](https://mlsanity.com/paper/2101.08122/) (2021)
- [Focus on the Positives: Self-Supervised Learning for Biodiversity Monitoring](https://mlsanity.com/paper/2108.06435/) (2021)
- [Self-supervised Pre-training with Hard Examples Improves Visual Representations](https://mlsanity.com/paper/2012.13493/) (2020)
- [Towards Efficient and Effective Self-Supervised Learning of Visual Representations](https://mlsanity.com/paper/2210.09866/) (2022)

## Abstract (arXiv)

Domain-specific variants of contrastive learning can construct positive pairs from two distinct in-domain images, while traditional methods just augment the same image twice. For example, we can form a positive pair from two satellite images showing the same location at different times. Ideally, this teaches the model to ignore changes caused by seasons, weather conditions or image acquisition artifacts. However, unlike in traditional contrastive methods, this can result in undesired positive pairs, since we form them without human supervision. For example, a positive pair might consist of one image before a disaster and one after. This could teach the model to ignore the differences between intact and damaged buildings, which might be what we want to detect in the downstream task. Similar to false negative pairs, this could impede model performance. Crucially, in this setting only parts of the images differ in relevant ways, while other parts remain similar. Surprisingly, we find that downstream semantic segmentation is either robust to such badly matched pairs or even benefits from them. The experiments are conducted on the remote sensing dataset xBD, and a synthetic segmentation dataset for which we have full control over the pairing conditions. As a result, practitioners can use these domain-specific contrastive methods without having to filter their positive pairs beforehand, or might even be encouraged to purposefully include such pairs in their pretraining dataset.