Object Detection Networks on Convolutional Feature Maps

By Shaoqing Ren, Kaiming He, Ross B. Girshick, X. Zhang. 2015

433 citations recorded by the source corpus.

Most object detectors contain two important components: a feature extractor and an object classifier. The feature extractor has rapidly evolved with significant research efforts leading to better deep convolutional architectures. The object classifier, however, has not received much attention and many recent systems (like SPPnet and Fast/Faster R-CNN) use simple multi-layer perceptrons. This paper demonstrates that carefully designing deep networks for object classification is just as important. We experiment with region-wise classifier networks that use shared, region-independent convolutional features. We call them “Networks on Convolutional feature maps” (NoCs). We discover that aside from deep feature maps, a deep and convolutional per-region classifier is of particular importance for object detection, whereas latest superior image classification models (such as ResNets and GoogLeNets) do not directly lead to good detection accuracy without using such a per-region classifier. We show by experiments that despite the effective ResNets and Faster R-CNN systems, the design of NoCs is an essential element for the 1st-place winning entries in ImageNet and MS COCO challenges 2015.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Object Detection via a Multi-region and Semantic Segmentation-Aware CNN Model (2015, methodology): “Contemporary to our work are the approaches of [29, 9, 28] that are also based on the SPP-Net framework.”
Group Normalization (2018, background): “We find that replacing the 2fc box head with 4conv1fc (similar to [48]) can better leverage GN.”
Fully Convolutional Instance-Aware Semantic Segmentation (2016, background): “As observed empirically, a considerably complex sub-network in the last step is necessary to obtain good accuracy [36, 9].”
You Only Look Once: Unified, Real-Time Object Detection (2015, background): “show that adding both convolutional and connected layers to pretrained networks can improve performance [28].”
R-FCN: Object Detection via Region-based Fully Convolutional Networks (2016, neutral): “Similar observations are reported in [20].”
MegDet: A Large Mini-Batch Object Detector (2017, methodology): “Based on our MegDet, we integrate the techniques including OHEM [35], atrous convolution [40, 2], stronger base models [38, 18], large kernel [28], segmentation supervision [27, 34], diverse network structure [12, 32, 36], contextual modules [22, 9], ROIAlign [14] and multi-scale training and testin”
Proposal-Free Network for Instance-Level Object Segmentation (2015, neutral): “Deep convolutional neural networks (DCNN) have achieved great success in object classification [33] [17] [31] [37], object detection [28] [26] [29] and object segmentation [23] [4] [14] [22].”
Towards Explainable Deep Neural Networks (xDNN) (2019, neutral): “According to [24], VGG–VD– 16 has a simple structure and it can achieve a better performance in comparison with other pre-trained deep neural networks.”
Training CNNs with Low-Rank Filters for Efficient Image Classification (2015, methodology): “Convolutional neural networks (CNNs) have been used increasingly succesfully to solve challenging computer vision problems such as image classification (Krizhevsky et al., 2012), object detection (Ren et al., 2015), and human pose estimation (Tompson et al., 2015).”
Action Recognition using Visual Attention (2015, background): “Convolutional Neural Networks (CNNs) have been highly successful in image classification and object recognition tasks (Ren et al., 2015; Wu et al., 2015).”
Accelerating Very Deep Convolutional Networks for Classification and Detection (2015, background): “Index Terms—Convolutional Neural Networks, Acceleration, Image Classification, Object Detection F”
Understanding Image Representations by Measuring Their Equivariance and Equivalence (2014, methodology): “This architectural difference makes ResNet slightly harder to use as a plug-in replacement for ANet in some applications (Ren et al. 2017), but the performance is generally far better than ANet and Vgg16.”
Pixel-Level Encoding and Depth Layering for Instance-Level Semantic Labeling (2016, methodology): “Driven by the success of deep learning based object detectors such as R-CNN [12] or its variants [11,24,25], recent methods rely on these detections for instance segmentation.”
HyperNet: Towards Accurate Region Proposal Generation and Joint Object Detection (2016, methodology): “Ren et al. [29] show that a ConvNet on pre-computed feature maps performs well.”
Reversible Recursive Instance-Level Object Segmentation (2015, background): “Recently, beyond the traditional object detection [7][24][5][14][4][25] and semantic segmentation tasks [1][18][2][32][16], instance-level object segmentation has attracted much attention [10][11][17][26][31][15].”
An Analysis of Scale Invariance in Object Detection - SNIP (2017, methodology): “In [33], an image pyramid was generated and maxout [12] was used to select features from a pair of scales closer to the resolution of the pre-trained dataset during inference.”