In Section 4.3, we demonstrate that Context Adjustment can improve pseudo-marks by 2.0% mean Intersection over Union on average and overall achieves a new state-of-the-art by 66.1% mIoU on the val set and 66.7% mIoU on the test set of PASCAL VOC 2012, and 33.4% mIoU on the val se...
We propose a self-supervised equivariant attention mechanism to narrow the supervision gap between fully and weakly supervised semantic segmentation by introducing additional self-supervision
We made a discovery that only a few labelled points is needed for existing point cloud encoder networks to produce very competitive performance for the point cloud segmentation task
Nuclei segmentation is a critical step in the automatic analysis of histopathology images, because the nuclear features such as average size, density and nucleusto-cytoplasm ratio are related to the clinical diagnosis and management of cancer
A two-stage framework called WS2 is proposed to overcome common challenges faced by many synthesize-refine scheme-based methods that are most successful in weakly-supervised video object segmentation
We propose an efficient end-to-end Intra-Class Discriminator approach, which dedicates to the intra-class discrimination between the foreground and the background pixels in each image
Our framework consists of a unary segmentation network to predict the class probability map, and a pairwise affinity network to learn affinity and refine the results of the unary network
MM '20: The 28th ACM International Conference on Multimedia
Seattle
WA
USA
..., pp.2085-2094, (2020)
In the weakly supervised segmentation task with only image-level labels, a common step in many existing algorithms is first to locate the image regions corresponding to each existing class with the Class Activation Maps (CAMs), and then generate the pseudo ground truth masks base...
european conference on computer vision, pp.571-587, (2020)
Learning semantic segmentation models requires a huge amount of pixel-wise labeling. However, labeled data may only be available abundantly in a domain different from the desired target domain, which only has minimal or no annotations. In this work, we propose a novel framework...
Following the same evaluation protocol from other competing approaches, we report mean average precision with four intersection over union thresholds, denoted by mAPkr where k denotes the different values of IoU and k = {0.25, 0.50, 0.70, 0.75}
We introduce Adaptive Pseudo Labeling for the positive pseudo labeling step, which dynamically selects the positive pseudo labels without the need of determining the threshold traditionally used in Pseudo Labeling
Experimental Results: The final results with the standard mean intersection over union criterion for weakly supervised semantic segmentation track of both LID19 and LID20 challenges are shown in Table 3
Since the ground truth labels are not available for sub-categories, we present visualizations of clustering results in Figure 5 to measure the quality, in which each parent class shows 3 example clusters
Point clouds provide intrinsic geometric information and surface context for scene understanding. Existing methods for point cloud segmentation require a large amount of fully labeled data. Using advanced depth sensors, collection of large scale 3D dataset is no longer a cumber...
We introduce a novel Multiple instance learning task based on a new kind of bag level label called unique class count, which is the number of unique classes or the number of clusters among all the instances inside the bag
We have used the SEN12MS dataset and the data provided in the frame of the IEEE-GRSS 2020 Data Fusion Contest to address the challenge of learning semantic segmentation models for global land cover mapping from inaccurate and inexact labels