CrowdHuman: A Benchmark for Detecting Human in a Crowd
arXiv:1805.00123
Abstract
Human detection has witnessed impressive progress in recent years. However, the occlusion issue of detecting human in highly crowded environments is far from solved. To make matters worse, crowd scenarios are still under-represented in current human detection benchmarks. In this paper, we introduce a new dataset, called CrowdHuman, to better evaluate detectors in crowd scenarios. The CrowdHuman dataset is large, rich-annotated and contains high diversity. There are a total of human instances from the train and validation subsets, and persons per image, with various kinds of occlusions in the dataset. Each human instance is annotated with a head bounding-box, human visible-region bounding-box and human full-body bounding-box. Baseline performance of state-of-the-art detection frameworks on CrowdHuman is presented. The cross-dataset generalization results of CrowdHuman dataset demonstrate state-of-the-art performance on previous dataset including Caltech-USA, CityPersons, and Brainwash without bells and whistles. We hope our dataset will serve as a solid baseline and help promote future research in human detection tasks.
References in corpus (2)
Cited by in corpus (67)
- TransTrack: Multiple Object Tracking with Transformer
- A Review and Comparative Study on Probabilistic Object Detection in Autonomous Driving
- From Handcrafted to Deep Features for Pedestrian Detection: A Survey
- Efficient DETR: Improving End-to-End Object Detector with Dense Prior
- SoccerNet-Tracking: Multiple Object Tracking Dataset and Benchmark in Soccer Videos
- ByteTrack: Multi-Object Tracking by Associating Every Detection Box
- Sparse R-CNN: End-to-End Object Detection with Learnable Proposals
- On the detection-to-track association for online multi-object tracking
- CrowdPose: Efficient Crowded Scenes Pose Estimation and A New Benchmark
- Bounding Box Regression with Uncertainty for Accurate Object Detection
- DETR for Crowd Pedestrian Detection
- FCOS: A simple and strong anchor-free object detector
- Semi-TCL: Semi-Supervised Track Contrastive Representation Learning
- NMS by Representative Region: Towards Crowded Pedestrian Detection by Proposal Pairing
- iQIYI-VID: A Large Dataset for Multi-modal Person Identification
- Contrastive Learning for Multi-Object Tracking with Transformers
- Adaptive NMS: Refining Pedestrian Detection in a Crowd
- Detection in Crowded Scenes: One Proposal, Multiple Predictions
- SurgiTrack: Fine-Grained Multi-Class Multi-Tool Tracking in Surgical Videos
- Tracking Objects as Points
- Human-Object Interaction Prediction in Videos through Gaze Following
- MOTR: End-to-End Multiple-Object Tracking with Transformer
- Adapted Center and Scale Prediction: More Stable and More Accurate
- Real-time 3D Deep Multi-Camera Tracking
- Analyzing Worldwide Social Distancing through Large-Scale Computer Vision
- End-to-End Object Detection with Fully Convolutional Network
- Pose2Seg: Detection Free Human Instance Segmentation
- Towards pedestrian head tracking: A benchmark dataset and a multi-source data fusion network
- Double Anchor R-CNN for Human Detection in a Crowd
- Monocular, One-stage, Regression of Multiple 3D People
- LLA: Loss-aware Label Assignment for Dense Pedestrian Detection
- YOLO11-JDE: Fast and Accurate Multi-Object Tracking with Self-Supervised Re-ID
- Pedestrian Motion State Estimation From 2D Pose
- Work-Efficient Parallel Non-Maximum Suppression Kernels
- PedHunter: Occlusion Robust Pedestrian Detector in Crowded Scenes
- Confluence: A Robust Non-IoU Alternative to Non-Maxima Suppression in Object Detection
- OTA: Optimal Transport Assignment for Object Detection
- Generalizable Pedestrian Detection: The Elephant In The Room
- Hashing-based Non-Maximum Suppression for Crowded Object Detection
- PS-RCNN: Detecting Secondary Human Instances in a Crowd via Primary Object Suppression
- V2F-Net: Explicit Decomposition of Occluded Pedestrian Detection
- Attribute-aware Pedestrian Detection in a Crowd
- Relational Learning for Joint Head and Human Detection
- SiamMOT: Siamese Multi-Object Tracking
- KinePose: A temporally optimized inverse kinematics technique for 6DOF human pose estimation with biomechanical constraints
- The Impact of Partial Occlusion on Pedestrian Detectability
- Looking Beyond Two Frames: End-to-End Multi-Object Tracking Using Spatial and Temporal Transformers
- NOH-NMS: Improving Pedestrian Detection by Nearby Objects Hallucination
- Focus on Semantic Consistency for Cross-domain Crowd Understanding
- Multiple Object Tracking with Correlation Learning
- iShape: A First Step Towards Irregular Shape Instance Segmentation
- Joint Learning Architecture for Multiple Object Tracking and Trajectory Forecasting
- Deep Mixture Density Network for Probabilistic Object Detection
- SADet: Learning An Efficient and Accurate Pedestrian Detector
- Simple multi-dataset detection
- Inception Convolution with Efficient Dilation Search
- Visible Feature Guidance for Crowd Pedestrian Detection
- Traversing the Subspace of Adversarial Patches
- Variational Pedestrian Detection
- Multi-object Tracking with Tracked Object Bounding Box Association
- Dense Scene Multiple Object Tracking with Box-Plane Matching
- A Top-down Approach to Articulated Human Pose Estimation and Tracking
- Automated Detection of Patients in Hospital Video Recordings
- Multi-Object Tracking with Hallucinated and Unlabeled Videos
- PANDA: A Gigapixel-level Human-centric Video Dataset
- Delving into the Imbalance of Positive Proposals in Two-stage Object Detection
- Working with scale: 2nd place solution to Product Detection in Densely Packed Scenes [Technical Report]