Objects as Points

By Xingyi Zhou, Dequan Wang, Philipp Krahenbuhl. 2019

3,766 citations recorded by the source corpus.

Detection identifies objects as axis-aligned boxes in an image. Most successful object detectors enumerate a nearly exhaustive list of potential object locations and classify each. This is wasteful, inefficient, and requires additional post-processing. In this paper, we take a different approach. We model an object as a single point --- the center point of its bounding box. Our detector uses keypoint estimation to find center points and regresses to all other object properties, such as size, 3D location, orientation, and even pose. Our center point based approach, CenterNet, is end-to-end differentiable, simpler, faster, and more accurate than corresponding bounding box based detectors. CenterNet achieves the best speed-accuracy trade-off on the MS COCO dataset, with 28.1% AP at 142 FPS, 37.4% AP at 52 FPS, and 45.1% AP with multi-scale testing at 1.4 FPS. We use the same approach to estimate 3D bounding box in the KITTI benchmark and human pose on the COCO keypoint dataset. Our method performs competitively with sophisticated multi-stage methods and runs in real-time.

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

End-to-End Object Detection with Transformers (2020, background): “More recent detectors [37,23,53] use non-unique assignment rules between ground truth and predictions together with an NMS.”
Bridging the Gap Between Anchor-Based and Anchor-Free Detection via Adaptive Training Sample Selection (2019, methodology): “[70] use keypoint estimation to find center point of objects and regress to all other properties including size, 3D location, orientation and pose.”
Pix2seq: A Language Modeling Framework for Object Detection (2021, methodology): “To predict a set of bounding boxes, architectures of modern detectors are speciﬁcally designed to produce a large set of proposals (Girshick, 2015; Ren et al., 2015; Cai & Vasconcelos, 2018), anchors (Lin et al., 2017b), or window centers (Tian et al., 2019; Zhou et al., 2019).”
Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection (2023, neutral): “We use the notations “CC”, “SBU”, “VG”, “OI”, “O365”, and “YFCC” for Conceptual Captions [45], SBU Captions [36], Visual Genome [20], OpenImage [22], Objects365 [63], YFCC100M [49] respectively.”
Ego4D: Around the World in 3,000 Hours of Egocentric Video (2021, background): “[241] Xingyi Zhou, Dequan Wang, and Philipp Krähenbühl.”
A Comprehensive Survey of Continual Learning: Theory, Method and Application (2023, methodology): “The idea of knowledge distillation is then introduced to other detection frameworks, such as CenterNet [508] (SID [328]), RetinaNet [259] (RILOD [247]), GFLV1 [250] (ERD [114]) and Faster R-CNN [358] (CIFRCN [152], Faster ILOD [327], DMC [493], BNC [96], IOD-ML [222]).”
EfficientDet: Scalable and Efficient Object Detection (2019, background): “Recently, one-stage detectors have attracted substantial attention due to their efﬁciency and simplicity [14, 34, 36].”
EPicker is an exemplar-based continual learning approach for knowledge accumulation in cryoEM particle picking (2022, methodology): “CenterNet 29 was used as the basic network for particle picking in EPicker.”
Overcoming Catastrophic Forgetting in Incremental Object Detection via Elastic Response Distillation (2022, methodology): “[28] proposed SID approach for IOD on anchor-free detector and conducted experiments on FCOS [33] and CenterNet [39].”
Recent Advances of Continual Learning in Computer Vision: An Overview (2021, positive): “In object detection, Perez et al. [81] proposed OpeNended Centre nEt (ONCE) which adapts CentreNet [83] detector towards the continual learning scenario where new tasks are registered with the help of meta learning.”
Incremental Few-Shot Instance Segmentation (2021, methodology): “Incremental few-shot object detection has been considered in ONCE [23], which uses CenterNet [40] as a back-bone to learn a class-agnostic feature extractor and a per-class code generator network for novel classes.”
SID: Incremental Learning for Anchor-Free Object Detection via Selective and Inter-Related Distillation (2020, methodology): “We build our method on top of FCOS (Tian et al., 2019) and CenterNet (Zhou et al., 2019) using their public implementations.”
Incremental Few-Shot Object Detection (2020, positive): “Our ONCE method is based on the one-stage CentreNet [56] which is chosen because of its efficiency and competitive detection accuracy, as well as the fact that it can be easily decomposed into class-generic and specific parts for adaptation to the Incremental Few-Shot Detection problem.”
Vision Transformers for Dense Prediction (2021, background): “Fully-convolutional networks [35, 36] are the prototypical architecture for pixel-level dense prediction tasks such as semantic segmentation [7, 25, 55], monocular depth estimation [12, 16, 32], and keypoint detection [21, 57].”
CenterNet: Keypoint Triplets for Object Detection (2019, methodology): “keypoint-based object detection approaches [43, 29, 50, 49] are proposed to avoid the disadvantages of the use of anchor boxes and bounding box regression.”
DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR (2022, neutral): “Most classical detectors are anchor-based, using either anchor boxes (Ren et al., 2017; Girshick, 2015; Sun et al., 2021) or anchor points (Tian et al., 2019; Zhou et al., 2019).”
Conditional DETR for Fast Training Convergence (2021, methodology): “Typical methods include YOLOv1 [30], CornerNet [21], ExtremeNet [50], CenterNet [49, 6], FCOS [39] and others [23, 28, 52, 19, 51, 22, 15, 46, 47].”
FCOS3D: Fully Convolutional One-Stage Monocular 3D Object Detection (2021, methodology): “On the validation set, we compare our method with the best open-source detector, CenterNet.”