CBNet: A Novel Composite Backbone Network Architecture for Object Detection

Liu, Yudong; Wang, Yongtao; Wang, Siwei; Liang, Tingting; Zhao, Qijie; Tang, Zhi; Ling, Haibin

doi:10.1609/aaai.v34i07.6834

Cited by 206 publications

(119 citation statements)

References 18 publications

(44 reference statements)

Supporting

Mentioning

104

Contrasting

Order By: Relevance

“…NASNet [30] and EfficentNet [31] are also proposed to construct more effective backbone. Recently, Liu et al proposed CBNet [32], which adopts multi-branch parallel backbone with iteratively feeding the output features of the previous backbone to improve performance. However, all these feature extraction backbones not only afford a large quantity of parameters that make model size burdensome, but also cause the high computational cost.…”

Section: B From Common Backbones To Lightweight Backbonesmentioning

confidence: 99%

TIB-Net: Drone Detection Network With Tiny Iterative Backbone

et al. 2020

View full text Add to dashboard Cite

With the widespread application of drone in commercial and industrial fields, drone detection has received increasing attention in public safety and others. However, due to various appearance of smallsize drones, changeable and complex environments, and limited memory resources of edge computing devices, drone detection remains a challenging task nowadays. Although deep convolutional neural network (CNN) has shown powerful performance in object detection in recent years, most existing CNN-based methods cannot balance detection performance and model size well. To solve the problem, we develop a drone detection network with tiny iterative backbone named TIB-Net. In this network, we propose a structure called cyclic pathway, which enhances the capability to extract effective features of small object, and integrate it into existing efficient method Extremely Tiny Face Detector (EXTD). This method not only significantly improves the accuracy of drone detection, but also keeps the model size at an acceptable level. Furthermore, we integrate spatial attention module into our network backbone to emphasize information of small object, which can better locate small-size drone and further improve detection performance. In addition, we present massive manual annotations of object bounding boxes for our collected 2860 drone images as a drone benchmark dataset, which is now publicly available 1. In this work, we conduct a series of experiments on our collected dataset to evaluate TIB-Net, and the result shows that our proposed method achieves mean average precision of 89.2% with model size of 697.0KB, which achieves the state-of-the-art results compared with existing methods.

show abstract

Section: B From Common Backbones To Lightweight Backbonesmentioning

confidence: 99%

TIB-Net: Drone Detection Network With Tiny Iterative Backbone

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Additionally, fusion operation focuses on two aspects, aggregation path, and expansion path. For the former, PANet [27] adds an extraction bottom-up path and CBNet [28] overlays parallel feature maps in different size. The latter outperforms the former for defect classes in detection, since CBNet possesses more parameters and more aggregating feature, as illustrated in Table 3.…”

Section: Neck For Feature Integrating and Refiningmentioning

confidence: 99%

Automatic Fabric Defect Detection Using Cascaded Mixed Feature Pyramid with Guided Localization

Zhang

2020

Sensors

View full text Add to dashboard Cite

Generic object detection algorithms for natural images have been proven to have excellent performance. In this paper, fabric defect detection on optical image datasets is systematically studied. In contrast to generic datasets, defect images are multi-scale, noise-filled, and blurred. Back-light intensity would also be sensitive for visual perception. Large-scale fabric defect datasets are collected, selected, and employed to fulfill the requirements of detection in industrial practice in order to address these imbalanced issues. An improved two-stage defect detector is constructed for achieving better generalization. Stacked feature pyramid networks are set up to aggregate cross-scale defect patterns on interpolating mixed depth-wise block in stage one. By sharing feature maps, center-ness and shape branches merges cascaded modules with deformable convolution to filter and refine the proposed guided anchors. After balanced sampling, the proposals are down-sampled by position-sensitive pooling for region of interest, in order to characterize interactions among fabric defect images in stage two. The experiments show that the end-to-end architecture improves the occluded defect performance of region-based object detectors as compared with the current detectors.

show abstract

“…Group multi-scale feature fusion part: a composition method of backbone network for object detection was presented in [26]. Inspired by this work, we propose group multi-scale feature fusion for HSI SR. After a 3D convolution, the feature maps are divided into N groups along the channel dimension.…”

Section: B Architecture Of Mffa-3d Blockmentioning

confidence: 99%

Hyperspectral Image Super Resolution Based on Multiscale Feature Fusion and Aggregation Network With 3-D Convolution

Tang

2020

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

The spectral resolution of hyperspectral images (HSIs) is very high. Nevertheless, their spatial resolution is low due to various hardware limitations. Therefore, it is important to study HSI super-resolution to improve their spatial resolution. In this paper, for hyperspectral single-image super-resolution, we propose a multi-scale feature fusion and aggregation network with 3D convolution (MFFA-3D) by cascading the MFFA-3D block. The MFFA-3D block includes group multi-scale feature fusion part and multi-scale feature aggregation part. In group multi-scale feature fusion part, a novel group multi-scale feature fusion method is proposed. Group feature fusion module and two-step multi-scale module are proposed in multi-scale feature aggregation part. In order to prevent spectral distortion, a spectral gradient loss function is proposed and combined with the mean square error (MSE) loss function to form the final loss function. Since the proposed super-resolution network is full 3D convolutional network, our method can perform direct super-resolution transfer even if the number of the bands of test images is different from that of training images. The experiments over simulated and real hyperspectral images demonstrate the superiority of the proposed method in terms of qualitative and quantitative evaluation.

show abstract

CBNet: A Novel Composite Backbone Network Architecture for Object Detection

Cited by 206 publications

References 18 publications

TIB-Net: Drone Detection Network With Tiny Iterative Backbone

TIB-Net: Drone Detection Network With Tiny Iterative Backbone

Automatic Fabric Defect Detection Using Cascaded Mixed Feature Pyramid with Guided Localization

Hyperspectral Image Super Resolution Based on Multiscale Feature Fusion and Aggregation Network With 3-D Convolution

Contact Info

Product

Resources

About