YOLOv4: Optimal Speed and Accuracy of Object Detection

Abstract

There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at https://github.com/AlexeyAB/darknet

References in corpus (8)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Improved Regularization of Convolutional Neural Networks with Cutout
• Searching for Activation Functions
• Random Erasing Data Augmentation
• Learning Spatial Fusion for Single-Shot Object Detection
• RetinaMask: Learning to predict masks improves state-of-the-art single-shot detection for free
• Label Refinement Network for Coarse-to-Fine Semantic Segmentation
• RDSNet: A New Deep Architecture for Reciprocal Object Detection and Instance Segmentation

Cited by in corpus (185)

• YOLOX: Exceeding YOLO Series in 2021
• All One Needs to Know about Metaverse: A Complete Survey on Technological Singularity, Virtual Ecosystem, and Research Agenda
• Real-Time Polyp Detection, Localization and Segmentation in Colonoscopy Using Deep Learning
• Edge YOLO: Real-Time Intelligent Object Detection System Based on Edge-Cloud Cooperation in Autonomous Vehicles
• PP-YOLO: An Effective and Efficient Implementation of Object Detector
• Probabilistic two-stage detection
• Scaled-YOLOv4: Scaling Cross Stage Partial Network
• TPH-YOLOv5: Improved YOLOv5 Based on Transformer Prediction Head for Object Detection on Drone-captured Scenarios
• Real-Time Accident Detection in Traffic Surveillance Using Deep Learning
• LittleYOLO-SPP: A Delicate Real-Time Vehicle Detection Algorithm
• PP-YOLOv2: A Practical Object Detector
• A Deep Learning Based Automatic Defect Analysis Framework for In-situ TEM Ion Irradiations
• Object Detection Using Sim2Real Domain Randomization for Robotic Applications
• Context Understanding in Computer Vision: A Survey
• EEEA-Net: An Early Exit Evolutionary Neural Architecture Search
• New SAR target recognition based on YOLO and very deep multi-canonical correlation analysis
• Performance Study of YOLOv5 and Faster R-CNN for Autonomous Navigation around Non-Cooperative Targets
• A Systematic Review on Computer Vision-Based Parking Lot Management Applied on Public Datasets
• TAU: A Framework for Video-Based Traffic Analytics Leveraging Artificial Intelligence and Unmanned Aerial Systems
• Learn an Effective Lip Reading Model without Pains
• A Transformer-based Generative Adversarial Network for Brain Tumor Segmentation
• On the detection-to-track association for online multi-object tracking
• A Vision-based Social Distancing and Critical Density Detection System for COVID-19
• Autonomous Driving with Deep Learning: A Survey of State-of-Art Technologies
• A Survey on Masked Facial Detection Methods and Datasets for Fighting Against COVID-19
• A Comprehensive Survey of Machine Learning Applied to Radar Signal Processing
• Unsupervised Cross-Modal Distillation for Thermal Infrared Tracking
• The Lottery Tickets Hypothesis for Supervised and Self-supervised Pre-training in Computer Vision Models
• An Online Semantic Mapping System for Extending and Enhancing Visual SLAM
• Dual Semantic Fusion Network for Video Object Detection
• Computer Stereo Vision for Autonomous Driving
• FoodLogoDet-1500: A Dataset for Large-Scale Food Logo Detection via Multi-Scale Feature Decoupling Network
• Simple Training Strategies and Model Scaling for Object Detection
• Instant-Teaching: An End-to-End Semi-Supervised Object Detection Framework
• Identification of Binary Neutron Star Mergers in Gravitational-Wave Data Using YOLO One-Shot Object Detection
• LSDNet: Trainable Modification of LSD Algorithm for Real-Time Line Segment Detection
• CSL-YOLO: A New Lightweight Object Detection System for Edge Computing
• Corner Proposal Network for Anchor-free, Two-stage Object Detection
• Object Hider: Adversarial Patch Attack Against Object Detectors
• Optimizing YOLOv7 for Semiconductor Defect Detection
• Implicit Feature Pyramid Network for Object Detection
• Self-supervised Learning for Panoptic Segmentation of Multiple Fruit Flower Species
• Deep Learning-Based Connector Detection for Robotized Assembly of Automotive Wire Harnesses
• Mask Detection and Classification in Thermal Face Images
• Group channel pruning and spatial attention distilling for object detection
• LPYOLO: Low Precision YOLO for Face Detection on FPGA
• Cooperating RPN's Improve Few-Shot Object Detection
• DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection
• Artificial and beneficial -- Exploiting artificial images for aerial vehicle detection
• Sparse Adversarial Attack to Object Detection
• Over-sampling De-occlusion Attention Network for Prohibited Items Detection in Noisy X-ray Images
• Domain Adaptive YOLO for One-Stage Cross-Domain Detection
• Where and What: Driver Attention-based Object Detection
• AReLU: Attention-based Rectified Linear Unit
• A Comparative Review of Recent Few-Shot Object Detection Algorithms
• 1st Place Solution of LVIS Challenge 2020: A Good Box is not a Guarantee of a Good Mask
• Control Distance IoU and Control Distance IoU Loss Function for Better Bounding Box Regression
• Rethinking the backbone architecture for tiny object detection
• VIPriors 1: Visual Inductive Priors for Data-Efficient Deep Learning Challenges
• Traffic Sign Detection With Event Cameras and DCNN
• A Free Lunch for Unsupervised Domain Adaptive Object Detection without Source Data
• Toward Zero-shot Character Recognition: A Gold Standard Dataset with Radical-level Annotations
• An Analysis of Deep Object Detectors For Diver Detection
• Leveraging Trajectory Prediction for Pedestrian Video Anomaly Detection
• Backdoor Attacks on Crowd Counting
• Efficient Golf Ball Detection and Tracking Based on Convolutional Neural Networks and Kalman Filter
• Deep Convolutional Neural Networks: A survey of the foundations, selected improvements, and some current applications
• Learning on Hardware: A Tutorial on Neural Network Accelerators and Co-Processors
• Panoptic-PolarNet: Proposal-free LiDAR Point Cloud Panoptic Segmentation
• Image-Based Vehicle Classification by Synergizing Features from Supervised and Self-Supervised Learning Paradigms
• Road Damage Detection and Classification with Detectron2 and Faster R-CNN
• FPB: Feature Pyramid Branch for Person Re-Identification
• Bridging Control-Centric and Data-Centric Optimization
• Scale-aware Automatic Augmentation for Object Detection
• Cycle Self-Training for Semi-Supervised Object Detection with Distribution Consistency Reweighting
• CrowdSim2: an Open Synthetic Benchmark for Object Detectors
• Physics Validation of Novel Convolutional 2D Architectures for Speeding Up High Energy Physics Simulations
• Spectro-Temporal RF Identification using Deep Learning
• You Only Look One-level Feature
• Workshop on Autonomous Driving at CVPR 2021: Technical Report for Streaming Perception Challenge
• AugNet: End-to-End Unsupervised Visual Representation Learning with Image Augmentation
• Are We There Yet? Learning to Localize in Embodied Instruction Following
• Virtual Underwater Datasets for Autonomous Inspections
• Automatic Detection of Cardiac Chambers Using an Attention-based YOLOv4 Framework from Four-chamber View of Fetal Echocardiography
• Towards Generalized and Incremental Few-Shot Object Detection
• Increasing FPS for single board computers and embedded computers in 2021 (Jetson nano and YOVOv4-tiny). Practice and review
• Detecting Environmental Violations with Satellite Imagery in Near Real Time: Land Application under the Clean Water Act
• Rock Hunting With Martian Machine Vision
• A Spacecraft Dataset for Detection, Segmentation and Parts Recognition
• Image-Text Pre-Training for Logo Recognition
• Small Object Detection for Near Real-Time Egocentric Perception in a Manual Assembly Scenario
• COLLIDE-PRED: Prediction of On-Road Collision From Surveillance Videos
• Parking Spot Classification based on surround view camera system
• 1st Place Solution to ICDAR 2021 RRC-ICTEXT End-to-end Text Spotting and Aesthetic Assessment on Integrated Circuit
• Instance Segmentation of Dislocations in TEM Images
• Improving Robustness of Learning-based Autonomous Steering Using Adversarial Images
• Fooling Object Detectors: Adversarial Attacks by Half-Neighbor Masks
• Achieving Real-Time LiDAR 3D Object Detection on a Mobile Device
• Computer Vision-based Social Distancing Surveillance Solution with Optional Automated Camera Calibration for Large Scale Deployment
• RiWNet: A moving object instance segmentation Network being Robust in adverse Weather conditions
• A High-Performance Adaptive Quantization Approach for Edge CNN Applications
• Lesion detection in contrast enhanced spectral mammography
• QFCNN: Quantum Fourier Convolutional Neural Network
• Infrastructure Node-based Vehicle Localization for Autonomous Driving
• Edge-Cloud Collaborated Object Detection via Difficult-Case Discriminator
• WAD: A Deep Reinforcement Learning Agent for Urban Autonomous Driving
• AFD-Net: Adaptive Fully-Dual Network for Few-Shot Object Detection
• A Real-time Low-cost Artificial Intelligence System for Autonomous Spraying in Palm Plantations
• Developing a Compressed Object Detection Model based on YOLOv4 for Deployment on Embedded GPU Platform of Autonomous System
• Pavement Distress Detection and Segmentation using YOLOv4 and DeepLabv3 on Pavements in the Philippines
• Crowdsourced-based Deep Convolutional Networks for Urban Flood Depth Mapping
• Autonomous Mosquito Habitat Detection Using Satellite Imagery and Convolutional Neural Networks for Disease Risk Mapping
• Exploring 2D Data Augmentation for 3D Monocular Object Detection
• RCLC: ROI-based joint conventional and learning video compression
• Evaluating Computer Vision Techniques for Urban Mobility on Large-Scale, Unconstrained Roads
• Semantic-Aware Environment Perception for Mobile Human-Robot Interaction
• RPATTACK: Refined Patch Attack on General Object Detectors
• Disentangle Your Dense Object Detector
• DPNET: Dual-Path Network for Efficient Object Detectioj with Lightweight Self-Attention
• EMDS-7: Environmental Microorganism Image Dataset Seventh Version for Multiple Object Detection Evaluation
• AinnoSeg: Panoramic Segmentation with High Perfomance
• Sequence-guided protein structure determination using graph convolutional and recurrent networks
• Vehicle Reconstruction and Texture Estimation Using Deep Implicit Semantic Template Mapping
• 3D Matting: A Benchmark Study on Soft Segmentation Method for Pulmonary Nodules Applied in Computed Tomography
• Training dataset generation for bridge game registration
• Comparison of Object Detection Algorithms Using Video and Thermal Images Collected from a UAS Platform: An Application of Drones in Traffic Management
• Vec2Instance: Parameterization for Deep Instance Segmentation
• A novel approach to remove foreign objects from chest X-ray images
• A Study on Evaluation Standard for Automatic Crack Detection Regard the Random Fractal
• 2nd Place Solution to ECCV 2020 VIPriors Object Detection Challenge
• Bio-inspired Robustness: A Review
• Boosting ship detection in SAR images with complementary pretraining techniques
• HAIR: Head-mounted AR Intention Recognition
• Machine learning for adjoint vector in aerodynamic shape optimization
• Learning Universal Shape Dictionary for Realtime Instance Segmentation
• DeepRepair: Style-Guided Repairing for DNNs in the Real-world Operational Environment
• Financial ticket intelligent recognition system based on deep learning
• ABCP: Automatic Block-wise and Channel-wise Network Pruning via Joint Search
• You Cannot Easily Catch Me: A Low-Detectable Adversarial Patch for Object Detectors
• A Deep Learning Approach for Masking Fetal Gender in Ultrasound Images
• Semantic Segmentation and Object Detection Towards Instance Segmentation: Breast Tumor Identification
• Improving Distinction between ASR Errors and Speech Disfluencies with Feature Space Interpolation
• An Experience Report on Machine Learning Reproducibility: Guidance for Practitioners and TensorFlow Model Garden Contributors
• Compound Figure Separation of Biomedical Images with Side Loss
• Histogram of Cell Types: Deep Learning for Automated Bone Marrow Cytology
• Inverting and Understanding Object Detectors
• Dual-Modality Vehicle Anomaly Detection via Bilateral Trajectory Tracing
• A Joint Intensity-Neuromorphic Event Imaging System for Resource Constrained Devices
• Pulmonary embolism identification in computerized tomography pulmonary angiography scans with deep learning technologies in COVID-19 patients
• SuperpixelGridCut, SuperpixelGridMean and SuperpixelGridMix Data Augmentation
• Out-of-Distribution Detection for Adaptive Computer Vision
• Objects as Extreme Points
• Scorpion detection and classification systems based on computer vision and deep learning for health security purposes
• A Framework for Real-time Traffic Trajectory Tracking, Speed Estimation, and Driver Behavior Calibration at Urban Intersections Using Virtual Traffic Lanes
• Achieving Real-Time Object Detection on MobileDevices with Neural Pruning Search
• Isotonic Data Augmentation for Knowledge Distillation
• Egocentric View Hand Action Recognition by Leveraging Hand Surface and Hand Grasp Type
• Adversarial Semantic Contour for Object Detection
• Automatic Camera Control and Directing with an Ultra-High-Definition Collaborative Recording System
• Memory-Efficient CNN Accelerator Based on Interlayer Feature Map Compression
• Towards Toxic and Narcotic Medication Detection with Rotated Object Detector
• NAS-FCOS: Efficient Search for Object Detection Architectures
• Incorporating Data Uncertainty in Object Tracking Algorithms
• LVIS Challenge Track Technical Report 1st Place Solution: Distribution Balanced and Boundary Refinement for Large Vocabulary Instance Segmentation
• Video based real-time positional tracker
• Multi-regime analysis for computer vision-based traffic surveillance using a change-point detection algorithm
• FPAENet: Pneumonia Detection Network Based on Feature Pyramid Attention Enhancement
• Proactive Interaction Framework for Intelligent Social Receptionist Robots
• Application of the Neural Network Dependability Kit in Real-World Environments
• Dense Multiscale Feature Fusion Pyramid Networks for Object Detection in UAV-Captured Images
• Object Detection-Based Variable Quantization Processing
• Research on Fast Text Recognition Method for Financial Ticket Image
• Metaknowledge Extraction Based on Multi-Modal Documents
• APEX-Net: Automatic Plot Extractor Network
• Anchor Distance for 3D Multi-Object Distance Estimation from 2D Single Shot
• An Adaptive Video Acquisition Scheme for Object Tracking and its Performance Optimization
• Auto-Detection of Tibial Plateau Angle in Canine Radiographs Using a Deep Learning Approach
• A Practical Framework for ROI Detection in Medical Images -- a case study for hip detection in anteroposterior pelvic radiographs
• VSS: A Storage System for Video Analytics [Technical Report]
• Research on Optimization Method of Multi-scale Fish Target Fast Detection Network
• ViFiT: Reconstructing Vision Trajectories from IMU and Wi-Fi Fine Time Measurements
• CelebHair: A New Large-Scale Dataset for Hairstyle Recommendation based on CelebA
• Self-supervised Learning of 3D Object Understanding by Data Association and Landmark Estimation for Image Sequence
• Actor-Action Video Classification CSC 249/449 Spring 2020 Challenge Report
• Curiosity-driven Intuitive Physics Learning