Video target detection method, system and device based on category external memory
A target detection and category technology, applied in the field of computer vision and pattern recognition, can solve the problems of auxiliary frame performance degradation, achieve the effect of enhancing robustness and discrimination, and improving accuracy
Active Publication Date: 2021-08-13
INST OF AUTOMATION CHINESE ACAD OF SCI
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0003] In order to solve the above-mentioned problems in the prior art, that is, the performance degradation of the prior art image recognition method based on feature aggregation is obvious when the number of auxiliary frames is small, the first aspect of the present invention provides a video recognition method based on category external memory Target detection method, described video target detection method comprises the following steps:
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
no. 1 example
[0076] Step B20, for each normalized first frame sequence in the normalized first frame sequence set, randomly select one frame as a training frame, m frames as auxiliary frames, m is a natural number, preferably to balance training speed and model For the selection of performance, select two frames as the auxiliary frames corresponding to the training frame, and select other numbers of frames to achieve similar effects. There is no specific limitation here, and the first frame corresponding to each frame of image is extracted through a general-purpose target detection network based on deep learning. an instance feature;
[0077] In one embodiment of the present invention, the general target detection network selects Faster R-CNN, and in other embodiments, an appropriate network can also be selected according to needs, and the present invention will not be described in detail here;
[0078] Step B30, input the features of the first instance into the soft-max pre-classifier to ...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
The invention belongs to the technical field, and in particular relates to a video target detection method, system and device based on category external memory, aiming to solve the problem in the prior art that the target detection performance drops significantly when the number of auxiliary frames is small. The invention includes: first training the video target detection model through the self-attention mechanism according to the training video information, then obtaining the enhanced instance features of the video to be tested through the trained video object detection model and the self-attention mechanism, and finally inputting the enhanced instance features into the general The classification branch and the bounding box regression branch of the object detection network obtain the object detection results. The invention reduces the sensitivity of the feature integration-based video target detection method in the prior art to the number of auxiliary frames, so that the target detection can be accurately performed even when there are few or no auxiliary frames.
Description
technical field [0001] The invention belongs to the field of computer vision and pattern recognition, and in particular relates to a video target detection method, system and device based on category external memory. Background technique [0002] Video object detection is an important and challenging computer vision task, which has a wide range of applications in security monitoring, intelligent video analysis, autonomous driving and other fields. However, it is not ideal to directly use image detectors to detect objects in videos due to motion blur, out-of-focus, etc. in some frames in the videos. Different from images, video data contains rich temporal motion information. Therefore, in order to solve the problem that image detectors do not perform well in low-quality frames in videos, many methods improve the performance of detectors by using temporal context information, such as A method based on feature aggregation. Although these methods have achieved a great improvem...
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G06K9/00G06K9/62G06N3/04G06N3/08G06F17/18
CPCG06N3/084G06F17/18G06V20/41G06V20/46G06V2201/07G06N3/045G06F18/2414
Inventor 张兆翔谭铁牛宋纯锋董文恺
Owner INST OF AUTOMATION CHINESE ACAD OF SCI