-
公开(公告)号：US20230186652A1
公开(公告)日：2023-06-15
申请号：US17662796
申请日：2022-05-10
Applicant: Anhui University
Inventor： Jie Chen , Haitao Wang , Bing Li , Zihan Cheng , Jingmin Xi , Yingjian Deng
CPC classification number: G06V20/597 , B60W40/09 , G06T7/10 , B60W2420/42 , G06T2207/20084 , G06T2207/30268
Abstract: The present disclosure provides a transformer-based driver distraction detection method and apparatus, belonging to the field of driving behavior analysis. The method includes: acquiring districted driving image data; building a driver distraction detection model FPT; inputting the acquired distracted driving image data into the driver distraction detection model FPT, analyzing the distracted driving image data by using the driver distraction detection model FPT, and determining a driver distraction state according to an analysis result. The present disclosure proposes a new network model, i.e., a driver distraction detection model FPT, based on Swin, Twins, and other models. Compared with a deep learning model, the FPT model compensates for the drawback that the deep learning model can only extract local features; the FPT model improves the classification accuracy and reduces the parameter quantity and calculation amount compared with the transformer model. The present disclosure adjusts the loss function of the whole network and adds label smoothing to the cross-entropy loss function, to increase the accuracy of classification, effectively suppress overfitting, and improve the detection accuracy.
-
2.发明公开公开(公告)号：US20230188671A1
公开(公告)日：2023-06-15
申请号：US17662210
申请日：2022-05-05
Applicant: Anhui University
Inventor： Jie Chen , Jianming Lv , Zihan Cheng , Zhixiang Huang , Haitao Wang , Bing Li , Huiyao Wan , Yun Feng
CPC classification number: H04N5/33 , G06V10/82 , G06V10/34 , G06V20/70 , G06T3/4038
Abstract: The present disclosure provides a fire source detection method and device under the condition of a small sample size, and a storage medium, and belongs to the field of target detection and industrial deployment. The method includes the steps of acquiring fire source image data from an industrial site; constructing a fire source detection model; inputting the fire source image data to the fire source detection model, and analyzing the fire source image data via the fire source detection model to obtain a detection result, where the detection result includes a specific location, precision and type of a fire source. By means of the method, the problems of insufficient sample capacity and difficulty in training under the condition of a small sample size are solved, and different enhancement methods are used to greatly increase the number and quality of samples and improve the over-fitting ability of models.
-
Patent Agency Ranking
3.发明公开公开(公告)号：US20230186436A1
公开(公告)日：2023-06-15
申请号：US17661177
申请日：2022-04-28
Applicant: Anhui University
Inventor： Jie Chen , Bing Li , Zihan Cheng , Haitao Wang , Jingmin Xi , Yingjian Deng
CPC classification number: B60W40/08 , G06V20/597 , G06V20/70 , G06V10/82 , G06T5/002 , G06F17/11 , B60W2040/0818 , B60W2420/42 , B60W2540/229
Abstract: The present disclosure provides a method for fine-grained detection of driver distraction based on unsupervised learning, belonging to the field of driving behavior analysis. The method includes: acquiring distracted driving image data; and inputting the acquired distracted driving image data into an unsupervised learning detection model, analyzing the distracted driving image data by using the unsupervised learning detection model, and determining a driver distraction state according to an analysis result. The unsupervised learning detection model includes a backbone network, projection heads, and a loss function; the backbone network is a RepMLP network structure incorporating a multilayer perceptron (MLP); the projection heads are each an MLP incorporating a residual structure; and the loss function is a loss function based on contrastive learning and a stop-gradient strategy.
-
公开(公告)号：US20230184927A1
公开(公告)日：2023-06-15
申请号：US17662402
申请日：2022-05-06
Applicant: Anhui University
Inventor： Jie Chen , Runfan Xia , Zhixiang Huang , Huiyao Wan , Xiaoping Liu , Zihan Cheng , Bocai Wu , Baidong Yao , Zheng Zhou , Jianming Lv , Yun Feng , Wentian Du , Jingqian Yu
IPC: G01S13/90 , G06T7/73 , G06V10/82 , G06V10/40 , G06V10/77 , G06T3/40 , G06V10/25 , G06V10/80 , G06V10/22 , G06T7/60 , G06V10/766 , G06V10/764
CPC classification number: G01S13/9027 , G06T7/73 , G06V10/82 , G06V10/40 , G06V10/7715 , G06T3/4046 , G06V10/25 , G06V10/806 , G06V10/225 , G06T7/60 , G06V10/766 , G06V10/764 , G06V2201/07 , G06T2207/20081 , G06T2207/20084 , G06T2207/10044
Abstract: A contextual visual-based synthetic-aperture radar (SAR) target detection method and apparatus, and a storage medium, belonging to the field of target detection is described. The method includes: obtaining an SAR image; and inputting the SAR image into a target detection model, and positioning and recognizing a target in the SAR image by using the target detection model, to obtain a detection result. In the present disclosure, a two-way multi-scale connection operation is enhanced through top-down and bottom-up attention, to guide learning of dynamic attention matrices and enhance feature interaction under different resolutions. The model can extract the multi-scale target feature information with higher accuracy, for bounding box regression and classification, to suppress interfering background information, thereby enhancing the visual expressiveness. After the attention enhancement module is added, the detection performance can be greatly improved with almost no increase in the parameter amount and calculation amount of the whole neck.
-
公开(公告)号：US12087046B2
公开(公告)日：2024-09-10
申请号：US17661177
申请日：2022-04-28
Applicant: Anhui University
Inventor： Jie Chen , Bing Li , Zihan Cheng , Haitao Wang , Jingmin Xi , Yingjian Deng
CPC classification number: G06V10/82 , G06V20/597 , G06V20/70
Abstract: The present disclosure provides a method for fine-grained detection of driver distraction based on unsupervised learning, belonging to the field of driving behavior analysis. The method includes: acquiring distracted driving image data; and inputting the acquired distracted driving image data into an unsupervised learning detection model, analyzing the distracted driving image data by using the unsupervised learning detection model, and determining a driver distraction state according to an analysis result. The unsupervised learning detection model includes a backbone network, projection heads, and a loss function; the backbone network is a RepMLP network structure incorporating a multilayer perceptron (MLP); the projection heads are each an MLP incorporating a residual structure; and the loss function is a loss function based on contrastive learning and a stop-gradient strategy.
-
公开(公告)号：US12056940B2
公开(公告)日：2024-08-06
申请号：US17662796
申请日：2022-05-10
Applicant: Anhui University
Inventor： Jie Chen , Haitao Wang , Bing Li , Zihan Cheng , Jingmin Xi , Yingjian Deng
CPC classification number: G06V20/597 , B60W40/09 , G06T7/10 , B60W2420/403 , G06T2207/20084 , G06T2207/30268
Abstract: The present disclosure provides a transformer-based driver distraction detection method and apparatus, belonging to the field of driving behavior analysis. The method includes: acquiring districted driving image data; building a driver distraction detection model FPT; inputting the acquired distracted driving image data into the driver distraction detection model FPT, analyzing the distracted driving image data by using the driver distraction detection model FPT, and determining a driver distraction state according to an analysis result. The present disclosure proposes a new network model, i.e., a driver distraction detection model FPT, based on Swin, Twins, and other models. Compared with a deep learning model, the FPT model compensates for the drawback that the deep learning model can only extract local features; the FPT model improves the classification accuracy and reduces the parameter quantity and calculation amount compared with the transformer model. The present disclosure adjusts the loss function of the whole network and adds label smoothing to the cross-entropy loss function, to increase the accuracy of classification, effectively suppress overfitting, and improve the detection accuracy.
-
7.发明授权公开(公告)号：US11818493B2
公开(公告)日：2023-11-14
申请号：US17662210
申请日：2022-05-05
Applicant: Anhui University
Inventor： Jie Chen , Jianming Lv , Zihan Cheng , Zhixiang Huang , Haitao Wang , Bing Li , Huiyao Wan , Yun Feng
CPC classification number: H04N5/33 , G06T3/4038 , G06V10/34 , G06V10/82 , G06V20/70
Abstract: The present disclosure provides a fire source detection method and device under the condition of a small sample size, and a storage medium, and belongs to the field of target detection and industrial deployment. The method includes the steps of acquiring fire source image data from an industrial site; constructing a fire source detection model; inputting the fire source image data to the fire source detection model, and analyzing the fire source image data via the fire source detection model to obtain a detection result, where the detection result includes a specific location, precision and type of a fire source. By means of the method, the problems of insufficient sample capacity and difficulty in training under the condition of a small sample size are solved, and different enhancement methods are used to greatly increase the number and quality of samples and improve the over-fitting ability of models.
-
-
-
-
-
-
Search Results
7
records
(took 0.011 seconds)
