Efficient Object Localization Using Convolutional Networks

2014/11/16 by Jonathan Tompson, Ross Goroshin, Tompson, Jonathan +7 · 1 voice · 27 citations
Computer Science · #Advanced Neural Network Applications #Artificial intelligence #Cascade #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Computer vision #Convolutional neural network #Detector #FOS: Computer and information sciences #Hand Gesture Recognition Systems #Human Pose and Action Recognition #Machine learning #Object detection #Offset (computer science) #Pattern recognition (psychology) #Pooling #Pose #Variance (accounting) #cs.CV

paper · pdf · doi:10.48550/arxiv.1411.4280

published in arXiv (Cornell University) (Cornell University) · 8 pages with 1 page of citations

openalex publication_date 2014/11/16 · arxiv published 2014/11/16 · arxiv created 2015/06/09 · arxiv updated 2015/06/10 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

Recent state-of-the-art performance on human-body pose estimation has been achieved with Deep Convolutional Networks (ConvNets). Traditional ConvNet architectures include pooling and sub-sampling layers which reduce computational requirements, introduce invariance and prevent over-training. These benefits of pooling come at the cost of reduced localization accuracy. We introduce a novel architecture which includes an efficient `position refinement' model that is trained to estimate the joint offset location within a small region of the image. This refinement model is jointly trained in cascade with a state-of-the-art ConvNet model to achieve improved accuracy in human joint location estimation. We show that the variance of our detector approaches the variance of human annotations on the FLIC dataset and outperforms all existing approaches on the MPII-human-pose dataset.

Citations

• Improving neural networks by preventing co-adaptation of feature detectors
• Gradient-based learning applied to document recognition
• Depth Map Prediction from a Single Image using a Multi-Scale Deep Network

Cited by

• A Neural Network-Based Real-time Casing Collar Recognition System for Downhole Instruments
• PerNodeDrop: A Method Balancing Specialized Subnets and Regularization in Deep Neural Networks
• Stratified Knowledge-Density Super-Network for Scalable Vision Transformers
• 3D Cal: An Open-Source Software Library for Depth Reconstruction on Vision-Based Tactile Sensors
• WhaleVAD-BPN: Improving Baleen Whale Call Detection with Boundary Proposal Networks and Post-processing Optimisation
• Convergence, design and training of continuous-time dropout as a random batch method
• Deep Neural Networks Inspired by Differential Equations
• High-Rate Mixout: Revisiting Mixout for Robust Domain Generalization
• U-Mamba2-SSL for Semi-Supervised Tooth and Pulp Segmentation in CBCT
• Dynamical Modeling of Behaviorally Relevant Spatiotemporal Patterns in Neural Imaging Data
• Self Identity Mapping
• A proposal for automated turbulence modelling
• Invariant Features for Global Crop Type Classification
• Streamlining the Development of Active Learning Methods in Real-World Object Detection
• Deep Neural Network Calibration by Reducing Classifier Shift with Stochastic Masking
• 3D Skeleton-Based Action Recognition: A Review
• S3CE-Net: Spike-guided Spatiotemporal Semantic Coupling and Expansion Network for Long Sequence Event Re-Identification
• CodeJeNN: A simple C++ neural network generator for physics applications
• Rethinking Uncertainty Quantification and Entanglement in Image Segmentation
• How Effective Can Dropout Be in Multiple Instance Learning ?
• Predicting Localized Primordial Star Formation with Deep Convolutional Neural Networks
• AttentionDrop: A Novel Regularization Method for Transformer Models

Discussions

• Efficient Object Localization Using Convolutional Networks [hn, 3 points, 0 comments]

Related

• Learning Human Pose Estimation Features with Convolutional Networks
• Composite Localization for Human Pose Estimation
• 2D Human Pose Estimation: A Survey
• Human Pose Estimation using Global and Local Normalization
• Human Pose Estimation from Depth Images via Inference Embedded Multi-task Learning