DSOD: Learning Deeply Supervised Object Detectors from Scratch

Abstract

We present Deeply Supervised Object Detector (DSOD), a framework that can learn object detectors from scratch. State-of-the-art object objectors rely heavily on the off-the-shelf networks pre-trained on large-scale classification datasets like ImageNet, which incurs learning bias due to the difference on both the loss functions and the category distributions between classification and detection tasks. Model fine-tuning for the detection task could alleviate this bias to some extent but not fundamentally. Besides, transferring pre-trained models from classification to detection between discrepant domains is even more difficult (e.g. RGB to depth images). A better solution to tackle these two critical problems is to train object detectors from scratch, which motivates our proposed DSOD. Previous efforts in this direction mostly failed due to much more complicated loss functions and limited training data in object detection. In DSOD, we contribute a set of design principles for training object detectors from scratch. One of the key findings is that deep supervision, enabled by dense layer-wise connections, plays a critical role in learning a good detector. Combining with several other principles, we develop DSOD following the single-shot detection (SSD) framework. Experiments on PASCAL VOC 2007, 2012 and MS COCO datasets demonstrate that DSOD can achieve better results than the state-of-the-art solutions with much more compact models. For instance, DSOD outperforms SSD on all three benchmarks with real-time detection speed, while requires only 1/2 parameters to SSD and 1/10 parameters to Faster RCNN. Our code and models are available at: https://github.com/szq0214/DSOD .

arXiv commentICCV 2017. Code and models are available at: https://github.com/szq0214/DSOD

References in corpus (14)

• Deep Residual Learning for Image Recognition
• Caffe: Convolutional Architecture for Fast Feature Embedding
• Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs
• Multi-Scale Context Aggregation by Dilated Convolutions
• Going Deeper with Convolutions
• ParseNet: Looking Wider to See Better
• DeepID3: Face Recognition with Very Deep Neural Networks
• Fully Convolutional Networks for Semantic Segmentation
• Rich feature hierarchies for accurate object detection and semantic segmentation
• YOLO9000: Better, Faster, Stronger
• PVANET: Deep but Lightweight Neural Networks for Real-time Object Detection
• The One Hundred Layers Tiramisu: Fully Convolutional DenseNets for Semantic Segmentation
• Inside-Outside Net: Detecting Objects in Context with Skip Pooling and Recurrent Neural Networks
• Cross Modal Distillation for Supervision Transfer

Cited by in corpus (25)

• Deep Learning for Generic Object Detection: A Survey
• Bringing AI To Edge: From Deep Learning's Perspective
• Towards Accurate One-Stage Object Detection with AP-Loss
• Receptive Field Block Net for Accurate and Fast Object Detection
• Single-Shot Refinement Neural Network for Object Detection
• Tiny-DSOD: Lightweight Object Detection for Resource-Restricted Usages
• Rethinking ImageNet Pre-training
• Object Detection with Deep Learning: A Review
• MDSSD: Multi-scale Deconvolutional Single Shot Detector for Small Objects
• TANet: Robust 3D Object Detection from Point Clouds with Triple Attention
• Improving Object Detection from Scratch via Gated Feature Reuse
• Structure Inference Net: Object Detection Using Scene-Level Context and Instance-Level Relationships
• Deep Adaptive Proposal Network for Object Detection in Optical Remote Sensing Images
• Optimizing Video Object Detection via a Scale-Time Lattice
• Single-Shot Bidirectional Pyramid Networks for High-Quality Object Detection
• Learning Deep Representations with Probabilistic Knowledge Transfer
• ScratchDet: Training Single-Shot Object Detectors from Scratch
• Rethinking Training from Scratch for Object Detection
• Fire SSD: Wide Fire Modules based Single Shot Detector on Edge Device
• Learning from Web Data: the Benefit of Unsupervised Object Localization
• Triply Supervised Decoder Networks for Joint Detection and Segmentation
• Deep Feature Pyramid Reconfiguration for Object Detection
• Composite Binary Decomposition Networks
• TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection
• Densely Semantic Enhancement for Domain Adaptive Region-free Detectors