CenterMask: Real-Time Anchor-Free Instance Segmentation

By Youngwan Lee, Jongyoul Park. Computer Vision and Pattern Recognition, 2019

666 citations recorded by the source corpus.

We propose a simple yet efficient anchor-free instance segmentation, called CenterMask, that adds a novel spatial attention-guided mask (SAG-Mask) branch to anchor-free one stage object detector (FCOS) in the same vein with Mask R-CNN. Plugged into the FCOS object detector, the SAG-Mask branch predicts a segmentation mask on each box with the spatial attention map that helps to focus on informative pixels and suppress noise. We also present an improved backbone networks, VoVNetV2, with two effective strategies: (1) residual connection for alleviating the optimization problem of larger VoVNet \cite{lee2019energy} and (2) effective Squeeze-Excitation (eSE) dealing with the channel information loss problem of original SE. With SAG-Mask and VoVNetV2, we deign CenterMask and CenterMask-Lite that are targeted to large and small models, respectively. Using the same ResNet-101-FPN backbone, CenterMask achieves 38.3%, surpassing all previous state-of-the-art methods while at a much faster speed. CenterMask-Lite also outperforms the state-of-the-art by large margins at over 35fps on Titan Xp. We hope that CenterMask and VoVNetV2 can serve as a solid baseline of real-time instance segmentation and backbone network for various vision tasks, respectively. The Code is available at https://github.com/youngwanLEE/CenterMask.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/226f98d49e73c58fdf21f87330d16ecfdbe6b3b0

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

Citations in the corpus

Cellpose 2.0: how to train your own model (2022, methodology): “appear to outperform the LiveCell model released with the LiveCell dataset [29], although this result should be interpreted cautiously since we evaluated performance after removing mask overlaps which the Cellpose model cannot handle correctly.”
Deep Occlusion-Aware Instance Segmentation with Overlapping BiLayers (2021, methodology): “Following typical models [21, 33] for instance segmentation, our model has three parts: (1) Backbone [22] with FPN [38] for ROI feature extraction; (2) Object detection head in charge of predicting bounding boxes as instance proposals.”
PETRv2: A Unified Framework for 3D Perception from Multi-Camera Images (2022, methodology): “Backbone Drive Lane Vehicle Lift-Splat [33] Res-101 0.729 0.200 0.321 FIERY [10] Res-101 - - 0.382 M 2 BEV [40] X-101 0.759 0.380 - BEVFormer [20] Res-101 0.801 0.257 0.448 PETRv2 Res-101 0.833 0.448 0.434 PETRv2 ∗ V2-99 0.856 0.490 0.463 Table 4.”
FCOS: A Simple and Strong Anchor-Free Object Detector (2020, methodology): “]. More importantly, due to its simple design, FCOS can be easily extended to solve other instance-level recognition tasks with minimal modiﬁcation, as already evidenced by instance segmentation [2], [20], [47], [51], keypoint detection [39], text spotting [26], and tracking [13], [44]. We expect to”
YOLOv4: Optimal Speed and Accuracy of Object Detection (2020, neutral): “CenterMask: Real-Time Anchor-Free Instance Segmentation [40] CenterMask-Lite MobileNetV2-FPN 600× 50.”
Is Pseudo-Lidar needed for Monocular 3D Object detection? (2021, methodology): “Note that the ﬁnal resolution of FPN features derived from DLA-34 and V2-99 network are different, strides=8, 16, 32, 64, 128 for DLA-34, strides= 4, 8, 16, 32, 64 for V2-99.”
Scaling Wide Residual Networks for Panoptic Segmentation (2020, methodology): “Building on top of WR-41, we further incorporate the simplified Squeeze-and-Excitation (SE) module [36, 49] (where only one fully connected layer is used), and the Switchable Atrous Convolution (SAC) [69], forming our baseline model.”
A survey on instance segmentation: state of the art (2020, neutral): “9] 42.0% 2018 SOLOv2 [134] 41.7% 2020 GCNet [130] 41.5% 2019 BlendMask [135] 41.3% 2020 SOLO [136] 40.4% 2019 Non-local Neural Networks [113] 40.3% 2017 Mask Scoring R-CNN [133] 39.6% 2019 CenterMask [137] 38.3% 2019 MaskLab+ [22] 38.1% 2017 TensorMask [91] 37.3% 2019 Mask R-CNN [67] 37.1% 2017 Pola”
Designing Network Design Strategies Through Gradient Path Analysis (2022, background): “Also, when VoVNet is stacked to 99 layers, its accuracy is even much lower than that of VoVNet-39 [18].”
Group DETR: Fast DETR Training with Group-Wise One-to-Many Assignment (2022, methodology): “We train these experiments for 24 epochs with VoVNetV2 [27] as the backbone and with the image size of 800 × 320.”
Vision-based robotic grasping from object localization, object pose estimation to grasp estimation for parallel grippers: a review (2020, background): “CenterMask [Lee and Park, 2020] adds a novel spatial attention-guided mask (SAG-Mask) branch to anchor-free one stage object detector (FCOS [Tian et al.”
Learning What Not to Segment: A New Perspective on Few-Shot Segmentation (2022, background): “However, collecting sufficient labeled data is notoriously time-consuming and labor-intensive, especially for dense prediction tasks, such as instance segmentation [2,3,15,21,59] and semantic segmentation [1,25,35,40,45].”
PETR: Position Embedding Transformation for Multi-View 3D Object Detection (2022, methodology): “When using the external data, PETR with VOVNetV2 [23] backbone achieves 50.4% NDS and 44.1% mAP.”