Mask-guided sample selection for semi-supervised instance segmentation
- PDF / 5,252,820 Bytes
- 19 Pages / 439.642 x 666.49 pts Page_size
- 85 Downloads / 209 Views
Mask-guided sample selection for semi-supervised instance segmentation Miriam Bellver1 · Amaia Salvador2 · Jordi Torres1 · Xavier Giro-i-Nieto2 Received: 31 October 2019 / Revised: 4 June 2020 / Accepted: 24 June 2020 / © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract Image segmentation methods are usually trained with pixel-level annotations, which require significant human effort to collect. Weakly-supervised pipelines are the most common solution to address this constraint because they are trained with lower forms of supervision, such as bounding boxes or scribbles. Semi-supervised methods are another option, that leverage a large amount of unlabeled data and a limited number of strongly-labeled samples. In this second setup, samples to be strongly-annotated can be selected randomly or with an active learning mechanism that chooses the ones that will maximize the model performance. In this work, we propose a sample selection approach to decide which samples to annotate for semi-supervised instance segmentation. Our method consists in first predicting pseudomasks for the unlabeled pool of samples, together with a score predicting the quality of each mask. This score is an estimate of the Intersection Over Union (IoU) of the segment with the ground truth mask. We study which samples should be annotated based on the quality score, leading to an improved performance for semi-supervised instance segmentation with low annotation budgets. Keywords Image segmentation · Semi-supervised learning · Active learning
Miriam Bellver
[email protected] Amaia Salvador [email protected] Jordi Torres [email protected] Xavier Giro-i-Nieto [email protected] 1
Barcelona Supercomputing Center (BSC), Jordi Girona Street, 29, 31, Barcelona, 08034, Spain
2
Universitat Polit`ecnica de Catalunya (UPC), Jordi Girona Street, 1-3, Barcelona, 08034, Spain
Multimedia Tools and Applications
1 Introduction Instance segmentation is a popular task in computer vision in which a mask and a class category are predicted for each target object in a given image. Typically, high-performing models rely on large datasets of annotated data, which are expensive to obtain. This work extends our previous study that presented a semi-supervised scheme, which we refer as BASIS [3] (from Budget Aware Semi-supervised semantic and Instance Segmentation). Given a low annotation budget, BASIS outperforms previous works on weakly or semisupervised semantic and instance segmentation. The BASIS approach for semi-supervised instance annotation is depicted in Fig. 1. The pipeline consists in firstly training an annotation network that uses only a few stronglyannotated samples. This annotation network is subsequently used to pseudo-annotate unlabeled or weakly-labeled samples. Later, a second segmentation network is trained with both the few strongly-annotated samples and the pseudo-annotations. In our previous solution [3], the subset of strongly-annotated samples was chosen randomly. In this work, we propose an alternativ
Data Loading...
