Multi-instance object segmentation with occlusion handling

By Yi-Ting Chen, Xiaokai Liu, Ming-Hsuan Yang. 2015

101 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Recurrent Neural Networks for Semantic Instance Segmentation (2017, methodology): “Our approach outperforms early proposal-based methods in [17, 8] by a significant margin across all IoU thresholds.”
Recurrent Instance Segmentation (2015, methodology): “A common problem with such piecewise learning methods is that each module does not learn to accommodate itself to the outputs of other modules.”
Pixelwise Instance Segmentation with a Dynamically Instantiated Network (2017, methodology): “Furthermore, since we are considering object detection outputs jointly with semantic segmentation predictions, we have some robustness to high-scoring false positive detections unlike methods such as [7, 19, 32] which refine object detections into segmentations.”
InstanceCut: From Edges to Instances with MultiCut (2016, background): “MCG [3] and then, in the second stage, a different network classifies these proposals [13, 22, 15, 11].”
The Cityscapes Dataset for Semantic Urban Scene Understanding (2016, methodology): “Either CNNs [23, 24] or non-parametric methods [10] are typically used, however, in both cases without coupling between individual predictions.”
Pixel-Level Encoding and Depth Layering for Instance-Level Semantic Labeling (2016, methodology): “Either the underlying region proposals, such as MCG [2], are directly used as instance segments [6,7,14], or the bounding boxes are refined to obtain instance masks [5,13].”
MaskLab: Instance Segmentation by Refining Object Detection with Semantic and Direction Features (2017, methodology): “Coupled with the mask proposals, SDS [28, 14] and CFM [17] incorporate mask-region features to improve the classification accuracy, while [29] exploit hypercolumn features (i.”
Semantic Instance Segmentation with a Discriminative Loss Function (2017, methodology): “[9] use MCG [3] to generate category-independent region proposals, followed by a classiﬁcation step.”
Recurrent Pixel Embedding for Instance Grouping (2017, background): “Alternately one can generate generic proposal segments and then label each one with a semantic detector [31, 12, 32, 16, 82, 34].”
Bottom-up Instance Segmentation using Deep Higher-Order CRFs (2016, methodology): “The SDS method of [13] was evaluated on the VOC Validation set by [4].”
ShapeMask: Learning to Segment Novel Objects by Refining Shape Priors (2019, methodology): “On the contrary, shapes are more informative (see right column) and have been used by numerous algorithms to help object segmentation [1, 47, 20, 7, 46].”
SAIL-VOS: Semantic Amodal Instance Level Video Object Segmentation – A Synthetic Dataset and Baselines (2019, background): “Explicit occlusion reasoning [43, 31, 74, 16] is related in that occluded objects are detected using bounding boxes or occlusion of speciﬁc objects are modeled, e.g ., people.”
Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers (2021, methodology): “[12] handle occlusion by incorporating category specific reasoning and exemplar-based shape prediction for instance segmentation.”
Semi-convolutional Operators for Instance Segmentation (2018, background): “In general, they implement a multi-stage pipeline that first generates region proposals or class agnostic boxes, and then classifies them [29,20,7,42,10,43,32].”
Multi-scale Patch Aggregation (MPA) for Simultaneous Detection and Segmentation (2016, methodology): “As shown in Table 5, proposal-based systems of [14, 15, 5, 3] take much longer time due to the high-quality proposal generation procedure [31].”
Learning Instance Occlusion for Panoptic Segmentation (2019, background): “eline. Occlusion ordering and layout learning. Occlusion handling is a long-studied computer vision task [53,12,47,20]. In the context of semantic segmentation, occlusion ordering has been adopted in [48,8]. [60] considers occlusion by way of amodal semantic segmentation, however, they annotate a sm”
Semantic Instance Segmentation via Deep Metric Learning (2017, methodology): “A common approach to this problem (e.g., [10, 4, 16, 6]) is ﬁrst to use some mechanism to predict object bounding boxes (e.g., by running a class-level object detector, or by using a class agnostic box proposal method such as EdgeBoxes), and then to run segmentation and classiﬁcation within each…”