Congju Du

Congju Du

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
17 November 2023 in Visual Intelligence
Verified
Journal Article
Open access
PDF file
BDBoyang Du
Boyang Du
CDCongju Du
Congju Du
LYLi Yu
Li Yu
Vehicle detection in dim light has always been a challenging task. In addition to the unavoidable noise, the uneven spatial distribution of light and dark due to vehicle lights and street lamps can further make the problem more difficult. Conventional image enhancement methods may produce over smoothing or over exposure problems, causing irreversible information loss to the vehicle targets to be subsequently detected. Therefore, we propose a multi-exposure generation and fusion network. In the multi-exposure generation network, we employ a single gated convolutional recurrent network with two-stream progressive exposure input to generate intermediate images with gradually increasing exposure, which are provided to the multi-exposure fusion network after a spatial attention mechanism. Then, a pre-trained vehicle detection model in normal light is used as the basis of the fusion network, and the two models are connected using the convolutional kernel channel dimension expansion technique. This allows the fusion module to provide vehicle detection information, which can be used to guide the generation network to fine-tune the parameters and thus complete end-to-end enhancement and training. By coupling the two parts, we can achieve detail interaction and feature fusion under different lighting conditions. Our experimental results demonstrate that our proposed method is better than the state-of-the-art detection methods after image luminance enhancement on the ODDS dataset.

21 May 2019 in Sensors
Verified
Journal Article
Open access
PDF file
CDCongju Du
Congju Du
BTBin Tang
Bin Tang
Radar unconventional active jamming, including unconventional deceptive jamming and barrage jamming, poses a serious threat to wideband radars. This paper proposes an unconventional-active-jamming recognition method for wideband radar. In this method, the visibility algorithm of converting the radar time series into graphs, called visibility graphs, is first given. Then, the visibility graph of the linear-frequency-modulation (LFM) signal is proved to be a regular graph, and the rationality of extracting features on visibility graphs is theoretically explained. Therefore, four features on visibility graphs, average degree, average clustering coefficient, Newman assortativity coefficient, and normalized network-structure entropy, are extracted from visibility graphs. Finally, a random-forests (RF) classifier is chosen for unconventional-active-jamming recognition. Experiment results show that recognition probability was over 90% when the jamming-to-noise ratio (JNR) was above 0 dB.

19 October 2022 in IEEE Transactions on Circuits and Systems for Video Technology
Verified
Journal Article
Closed access
CDCongju Du
Congju Du
ZYZengqiang Yan
Zengqiang Yan
HYHan Yu
Han Yu
LYLi Yu
Li Yu
ZXZixiang Xiong
Zixiang Xiong
Bottom-up human pose estimation decouples computational complexity from the number of people but requires additional operations to match the detected keypoints to each human instance. Existing approaches treat all keypoints equally while ignoring the relationships among keypoints, which in turn limit the performance ceilings. In this work, we propose a hierarchical associative encoding and decoding framework for bottom-up human pose estimation by introducing additional prior knowledge. Specifically, in addition to keypoint-level and instance-level associations, we further divide keypoints into groups and explore group-level associations. This way, prior knowledge is incorporated to determine the keypoint groups for better associative encoding. To deal with complex poses, we introduce a focal pulling loss to focus more on the hard-to-associate keypoints. Moreover, instead of using a pre-defined order for keypoint grouping, we propose a progressive associative decoding method to dynamically determine the order of keypoints for grouping, which helps reduce isolated keypoints. Experimental results on the MS-COCO, CrowdPose and MPII datasets show superior performance of our proposed associative encoding and decoding algorithms. More importantly, we prove, through validation, that hierarchical associative encoding and decoding can be used as a plug-n-play module for performance improvement regardless of backbone architecture. Our source code and pretrained models are available at https://github.com/ducongju/HAE.

7 June 2023 in Pattern Recognition Letters
Verified
Journal Article
Closed access
ZSZiqiao Shang
Ziqiao Shang
CDCongju Du
Congju Du
BLBingyin Li
Bingyin Li
ZYZengqiang Yan
Zengqiang Yan
LYLi Yu
Li Yu

10 July 2024 in Neural Networks
Verified
Journal Article
Closed access
CDCongju Du
Congju Du
ZYZengqiang Yan
Zengqiang Yan
ZXZixiang Xiong
Zixiang Xiong
LYLi Yu
Li Yu

3 January 2022 in IET Image Processing
Verified
Journal Article
Open access
PDF file
CDCongju Du
Congju Du
HYHan Yu
Han Yu
LYLi Yu
Li Yu
Multi‐person pose estimation is a challenging vision task that can be seriously affected by keypoint scale variation. Existing heatmap‐based approaches are devoted to reducing the effect by optimizing backbone architecture or loss functions, but the problem of an inaccurate heatmap representation with different keypoint scales still exists. A scale‐sensitive heatmap algorithm is presented to generate reasonable spatial and contextual features for the network to predict more precise coordinates, by systematically considering the standard deviation, truncated radius, and shape of Gaussian kernels. Specifically, the scale‐sensitive heatmap algorithm contains three parts: inter‐person heatmap, limited‐area heatmap, and shape‐aware heatmap. The inter‐person heatmap allocates different standard deviations for each human instance proportionally calculated by the keypoint‐based method, the limited‐area heatmap defines the truncated radius to limit the influence area of Gaussian kernels, and the shape‐aware heatmap modifies the Gaussian kernels generated by some ellipse‐shaped joints. Our scale‐sensitive heatmap algorithm outperforms the baseline by a considerable margin on the COCO and CrowdPose benchmark datasets. The code and pretrained models are available at https://github.com/ducongju/Scale‐sensitive‐Heatmap .