FreeAnchor: Learning to Match Anchors for Visual Object Detection
arXiv:1909.02466
Abstract
Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Unit (IoU). In this study, we propose a learning-to-match approach to break IoU restriction, allowing objects to match anchors in a flexible manner. Our approach, referred to as FreeAnchor, updates hand-crafted anchor assignment to "free" anchor matching by formulating detector training as a maximum likelihood estimation (MLE) procedure. FreeAnchor targets at learning features which best explain a class of objects in terms of both classification and localization. FreeAnchor is implemented by optimizing detection customized likelihood and can be fused with CNN-based detectors in a plug-and-play manner. Experiments on COCO demonstrate that FreeAnchor consistently outperforms their counterparts with significant margins.
arXiv commentAccepted by NeurIPS 2019
Cited by in corpus (64)
- YOLOv4: Optimal Speed and Accuracy of Object Detection
- YOLOX: Exceeding YOLO Series in 2021
- CSPNet: A New Backbone that can Enhance Learning Capability of CNN
- A Dataset And Benchmark Of Underwater Object Detection For Robot Picking
- PP-YOLO: An Effective and Efficient Implementation of Object Detector
- AutoAssign: Differentiable Label Assignment for Dense Object Detection
- Probabilistic two-stage detection
- Efficient DETR: Improving End-to-End Object Detector with Dense Prior
- A New Dataset, Poisson GAN and AquaNet for Underwater Object Grabbing
- Bridging the Gap Between Anchor-based and Anchor-free Detection via Adaptive Training Sample Selection
- SCPM-Net: An Anchor-free 3D Lung Nodule Detection Network using Sphere Representation and Center Points Matching
- Semantic-guided Pixel Sampling for Cloth-Changing Person Re-identification
- RepPoints V2: Verification Meets Regression for Object Detection
- Convolutional Neural Networks with Gated Recurrent Connections
- Object DGCNN: 3D Object Detection using Dynamic Graphs
- CBA: Contextual Background Attack against Optical Aerial Detection in the Physical World
- PDNet: Toward Better One-Stage Object Detection With Prediction Decoupling
- Dual-stream Network for Visual Recognition
- Dynamic Label Assignment for Object Detection by Combining Predicted IoUs and Anchor IoUs
- VarifocalNet: An IoU-aware Dense Object Detector
- AD-Cluster: Augmented Discriminative Clustering for Domain Adaptive Person Re-identification
- Dynamic R-CNN: Towards High Quality Object Detection via Dynamic Training
- A Ranking-based, Balanced Loss Function Unifying Classification and Localisation in Object Detection
- Prototype Mixture Models for Few-shot Semantic Segmentation
- Exploring Data Augmentation for Multi-Modality 3D Object Detection
- Corner Proposal Network for Anchor-free, Two-stage Object Detection
- Dynamic Scale Training for Object Detection
- Generalized Focal Loss V2: Learning Reliable Localization Quality Estimation for Dense Object Detection
- Dynamic Anchor Learning for Arbitrary-Oriented Object Detection
- SEED: Semantics Enhanced Encoder-Decoder Framework for Scene Text Recognition
- Side-Aware Boundary Localization for More Precise Object Detection
- Implicit Feature Pyramid Network for Object Detection
- Anchor Pruning for Object Detection
- TOOD: Task-aligned One-stage Object Detection
- Objects in Semantic Topology
- End-to-End Object Detection with Fully Convolutional Network
- Soft Anchor-Point Object Detection
- Oriented RepPoints for Aerial Object Detection
- LLA: Loss-aware Label Assignment for Dense Pedestrian Detection
- Beyond Max-Margin: Class Margin Equilibrium for Few-shot Object Detection
- Multiple instance active learning for object detection
- You Only Look One-level Feature
- OTA: Optimal Transport Assignment for Object Detection
- iffDetector: Inference-aware Feature Filtering for Object Detection
- Progressive Cluster Purification for Unsupervised Feature Learning
- MatrixNets: A New Scale and Aspect Ratio Aware Architecture for Object Detection
- Mask is All You Need: Rethinking Mask R-CNN for Dense and Arbitrary-Shaped Scene Text Detection
- Adaptive Class Suppression Loss for Long-Tail Object Detection
- Rethinking Performance Estimation in Neural Architecture Search
- IQDet: Instance-wise Quality Distribution Sampling for Object Detection
- Multiple Anchor Learning for Visual Object Detection
- Effective Fusion Factor in FPN for Tiny Object Detection
- The MIS Check-Dam Dataset for Object Detection and Instance Segmentation Tasks
- Disentangle Your Dense Object Detector
- Representation Sharing for Fast Object Detector Search and Beyond
- Probabilistic Ranking-Aware Ensembles for Enhanced Object Detections
- Variational Pedestrian Detection
- Which to Match? Selecting Consistent GT-Proposal Assignment for Pedestrian Detection
- Video Playback Rate Perception for Self-supervisedSpatio-Temporal Representation Learning
- Horizontal-to-Vertical Video Conversion
- A Holistically-Guided Decoder for Deep Representation Learning with Applications to Semantic Segmentation and Object Detection
- Location-Aware Box Reasoning for Anchor-Based Single-Shot Object Detection
- Learning Oriented Remote Sensing Object Detection via Naive Geometric Computing
- Anti-aliasing Semantic Reconstruction for Few-Shot Semantic Segmentation