Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Scalable Object Detection using Deep Neural Networks

Dumitru Erhan, Christian Szegedy, Alexander Toshev, Dragomir Anguelov

TL;DR

This work introduces DeepMultiBox, a scalable, class-agnostic object localization framework that predicts a fixed set of bounding boxes with confidences using a single deep network. The model learns via a novel assignment-based loss that jointly optimizes box coordinates and confidences by matching predictions to ground-truth boxes, enabling detection of multiple instances without classwise outputs. It demonstrates competitive performance on VOC2007 and ILSVRC2012 with only a small number of top proposals and a downstream classifier, highlighting strong scalability and cross-dataset transfer potential. The approach reduces reliance on class-specific detectors, enabling efficient detection across large label spaces while preserving the ability to generalize and detect multiple object instances.

Abstract

Deep convolutional neural networks have recently achieved state-of-the-art performance on a number of image recognition benchmarks, including the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC-2012). The winning model on the localization sub-task was a network that predicts a single bounding box and a confidence score for each object category in the image. Such a model captures the whole-image context around the objects but cannot handle multiple instances of the same object in the image without naively replicating the number of outputs for each instance. In this work, we propose a saliency-inspired neural network model for detection, which predicts a set of class-agnostic bounding boxes along with a single score for each box, corresponding to its likelihood of containing any object of interest. The model naturally handles a variable number of instances for each class and allows for cross-class generalization at the highest levels of the network. We are able to obtain competitive recognition performance on VOC2007 and ILSVRC2012, while using only the top few predicted locations in each image and a small number of neural network evaluations.

Scalable Object Detection using Deep Neural Networks

TL;DR

This work introduces DeepMultiBox, a scalable, class-agnostic object localization framework that predicts a fixed set of bounding boxes with confidences using a single deep network. The model learns via a novel assignment-based loss that jointly optimizes box coordinates and confidences by matching predictions to ground-truth boxes, enabling detection of multiple instances without classwise outputs. It demonstrates competitive performance on VOC2007 and ILSVRC2012 with only a small number of top proposals and a downstream classifier, highlighting strong scalability and cross-dataset transfer potential. The approach reduces reliance on class-specific detectors, enabling efficient detection across large label spaces while preserving the ability to generalize and detect multiple object instances.

Abstract

Deep convolutional neural networks have recently achieved state-of-the-art performance on a number of image recognition benchmarks, including the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC-2012). The winning model on the localization sub-task was a network that predicts a single bounding box and a confidence score for each object category in the image. Such a model captures the whole-image context around the objects but cannot handle multiple instances of the same object in the image without naively replicating the number of outputs for each instance. In this work, we propose a saliency-inspired neural network model for detection, which predicts a set of class-agnostic bounding boxes along with a single score for each box, corresponding to its likelihood of containing any object of interest. The model naturally handles a variable number of instances for each class and allows for cross-class generalization at the highest levels of the network. We are able to obtain competitive recognition performance on VOC2007 and ILSVRC2012, while using only the top few predicted locations in each image and a small number of neural network evaluations.

Paper Structure

This paper contains 17 sections, 5 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Previous work
  3. Proposed approach
  4. Model
  5. Training Objective
  6. Optimization
  7. Training Details
  8. Experimental results
  9. Network Architecture and Experiment Details
  10. VOC 2007
  11. Training methodology
  12. Evaluation methodology
  13. Discussion
  14. ILSVRC 2012 Detection Challenge
  15. Training methodology
  16. ...and 2 more sections

Figures (5)

  • Figure 1: Detection rate of class "object" vs number of bounding boxes per image. The model, used for these results, was trained on VOC 2012.
  • Figure 2: Sample of detection results on VOC 2007.
  • Figure 3: Precision-recall curves on selected VOC classes.
  • Figure 4: Some selected detection results on the ILSVRC-2012 detection challenge validation set.
  • Figure 5: Class-agnostic detection on ILSVRC-2012 (left) and VOC 2007 (right).