Amodal Instance Segmentation

By Ke Li, Jitendra Malik. 2016

184 citations recorded by the source corpus.

We consider the problem of amodal instance segmentation, the objective of which is to predict the region encompassing both visible and occluded parts of each object. Thus far, the lack of publicly available amodal segmentation annotations has stymied the development of amodal segmentation methods. In this paper, we sidestep this issue by relying solely on standard modal instance segmentation annotations to train our model. The result is a new method for amodal instance segmentation, which represents the first such method to the best of our knowledge. We demonstrate the proposed method’s effectiveness both qualitatively and quantitatively.

Subfields: Computer Vision and Pattern Recognition, Environmental Engineering, Geology

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

In Environmental Engineering

In Geology

Citations in the corpus

Hypersim: A Photorealistic Synthetic Dataset for Holistic Indoor Scene Understanding (2020, background): “For example, our mesh annotations provide segmentation information for occluded parts of the scene that aren’t visible in any image, which could be used for amodal segmentation problems (e.g., [42]).”
Panoptic-DeepLab: A Simple, Strong, and Fast Baseline for Bottom-Up Panoptic Segmentation (2019, background): “The former do not allow overlapping masks (thus creating occlusions among masks), while the latter allows overlaps, which might make the training target easier to optimize, similar to amodal segmentation [81, 41].”
Reconstructing Hand-Object Interactions in the Wild (2020, background): “Amodal instance masks, that include both visible and occluded parts of the object [23], are a good fit.”
3D-RCNN: Instance-Level 3D Object Reconstruction via Render-and-Compare (2018, background): “These entities are easier to learn, and ground-truth data is easy to obtain [30] or already available from real-world datasets like KITTI [13] and Pascal3D+ [58].”
Deep convolutional neural networks for mobile capture device-based crop disease classification in the wild (2019, background): “…object classiﬁcation and object detection (Russakovsky et al. (2015)), semantic segmentation (Shelhamer et al. (2016)), instance segmentation (Li and Malik (2016))), enabling the deﬁnition and addressing of new tasks that were not even possible up until the appear-95 ance of these techniques…”
Self-Supervised Scene De-Occlusion (2020, background): “[28] produces dummy supervision through pasting artificial occluders, while the absence of explicit ordering increases the difficulty when complicated occlusion relationship is present.”
3D Scene Graph: A Structure for Unified Semantics, 3D Space, and Camera (2019, background): “Since our semantic information resides in 3D space, we can infer the full extents of object occlusions without additional annotations and in a fully automatically way, considering the difficulties of data collection in previous works [29, 52, 16].”
Visual Concepts and Compositional Voting (2017, methodology): “It introduces difficulties for tasks such as object detection [16] or segmentation [20].”
Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers (2021, methodology): “Li and Malik [35] first propose a method by extending [34], which iteratively enlarges the modal bounding box following the direction of high heatmap values and synthetically adds occlusion.”
CityPersons: A Diverse Dataset for Pedestrian Detection (2017, background): “full extent of the persons bodies (“amodal bounding box” [22]).”
SAIL-VOS: Semantic Amodal Instance Level Video Object Segmentation – A Synthetic Dataset and Baselines (2019, methodology): “Amodal instance segmentation without amodal data was considered by Li and Malik [62].”
Semantic Amodal Segmentation (2015, methodology): “Next, inspired by Ke et al. [23] (which is not directly applicable to our setting), we propose a deep network we call ExpandMask .”
DeepVoting: An Explainable Framework for Semantic Part Detection under Partial Occlusion (2017, methodology): “Since there is no public occlusion dataset with semantic parts labelled, and it would be extremely difﬁcult and expensive to label the semantic parts which are truly occluded, we follow the similar procedure in (Li and Malik 2016) to synthesize an occlusion dataset.”