RECALL: Replay-based Continual Learning in Semantic Segmentation

By Andrea Maracani, Umberto Michieli, Marco Toldo, Pietro Zanuttigh. 2021

157 citations recorded by the source corpus.

Deep networks allow to obtain outstanding results in semantic segmentation, however they need to be trained in a single shot with a large amount of data. Continual learning settings where new classes are learned in incremental steps and previous training data is no longer available are challenging due to the catastrophic forgetting phenomenon. Existing approaches typically fail when several incremental steps are performed or in presence of a distribution shift of the background class. We tackle these issues by recreating no longer available data for the old classes and outlining a content inpainting scheme on the background class. We propose two sources for replay data. The first resorts to a generative adversarial network to sample from the class space of past learning steps. The second relies on web-crawled data to retrieve images containing examples of old classes from online databases. In both scenarios no samples of past steps are stored, thus avoiding privacy concerns. Replay data are then blended with new samples during the incremental steps. Our approach, RECALL, outperforms state-of-the-art methods.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging, Meta & Continual 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

In Computer Vision and Pattern Recognition

In Radiology, Nuclear Medicine and Imaging

In Meta & Continual Learning

Citations in the corpus

A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, methodology): “PLOP [292], RECALL [293], SSUL [294], Self-Training [295] and WILSON [296] explicitly apply the old model to generate pseudo-labels of the old classes.”
Towards Label-Efficient Incremental Learning: A Survey (2023, background): “However, for incremental dense prediction tasks such as image segmentation [Maracani et al. , 2021], the demand for labels is explosive.”
ALIFE: Adaptive Logit Regularizer and Feature Replay for Incremental Semantic Segmentation (2022, methodology): “Following the common practice in [2, 3, 6, 12, 13], we perform a cross-validation to choose other hyperparameters, e.”
Uncertainty-aware Contrastive Distillation for Incremental Semantic Segmentation (2022, methodology): “More recently, RECALL [41] employs a web crawler or a pre-trained conditional GAN to replay images for the past semantic classes.”
Representation Compensation Networks for Continual Semantic Segmentation (2022, background): “The knowledge can be stored by multiple types, like examples [4, 6, 9, 11, 73, 83], prototypes [35, 106, 107], generative networks [60], etc.”
Incremental Learning in Semantic Segmentation from Image Labels (2021, methodology): “A better solution is to incrementally add new classes to the pre-existing model, as done in some recent works [8, 21, 43, 45, 46].”
Endpoints Weight Fusion for Class Incremental Semantic Segmentation (2023, methodology): “It is worth noting that there are several methods [6, 34, 49] that introduce different forms of auxiliary data to assist continual semantic segmentation, helping the model build better pseudo-labels or enhance memory for old knowledge.”
CoMFormer: Continual Learning in Semantic and Panoptic Segmentation (2022, methodology): “Traditionally, it has been studied in the context of image classiﬁcation [6, 17, 19, 29, 33, 43, 45, 55] but there is a growing interest in its applications in segmentation [2–5,15,16,38–40,59].”
Few-Shot Incremental Learning for Label-to-Image Translation (2022, methodology): “To address the problem, reply-based methods [2, 4, 28, 38] retain old knowledge by rehearsing on previous training data.”
Deep Class-Incremental Learning: A Survey (2023, unknown): “Replay is a simple yet effective strategy, which has been widely applied to image-based camera localization [129], semantic segmentation [130], [131], [132], video classification [133], and action recognition [95].”
CoinSeg: Contrast Inter- and Intra- Class Representations for Incremental Segmentation (2023, background): “The historical data can also be obtained with web search engine [23] or generation models [27, 39, 41].”
Continual Semantic Segmentation Leveraging Image-level Labels and Rehearsal (2022, methodology): “, 2021] and 3) replaying past data stored in memory [Maracani et al., 2021].”
Class-Incremental Learning: A Survey (2023, background): “Besides, works also investigate the application of generative replay [160] and auxiliary data [160], [280] for data replay.”
Saving 100x Storage: Prototype Replay for Reconstructing Training Sample Distribution in Class-Incremental Semantic Segmentation (2023, background): “Among them, MiB [2], SDR [23], PLOP [9], SSUL [3], RCIL [31], RBC [33], DKD [1], UCD [29], and AWT [12] are non-replay-based, while RECALL [20], SSUL-M [3], and DKD-M [1] are replay-based methods.”