Target detection model training method and related device
A training method and target detection technology, applied in character and pattern recognition, instruments, computer components, etc., can solve problems such as the inability to use large-size or multi-size features and affect model performance, so as to improve model performance and reduce complexity Effect
Pending Publication Date: 2022-04-12
INSPUR BEIJING ELECTRONICS INFORMATION IND
View PDF0 Cites 1 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0002] In related technologies, although PIX2SEQ (a target detection architecture) converts the target detection task into a language model generation task, it still uses a transformer internally. Due to the characteristics of the transformer itself, when processing image features, it will calculate the current position and all other positions. The relationship between the complexity is proportional to the size of the feature map, making it impossible to use large-size or multi-size features, which in turn affects the performance of the model
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
specific Embodiment approach
[0130] Construct a sequence based on the real ground truth of the image, and do corresponding data enhancement. The ground truth of target detection has detection frames and corresponding categories. Due to different image sizes, the detection frames are also different. Therefore, it is necessary to unify the images to a uniform size, so that all detection frames can be quantified to a uniform scale. A small The vocabulary is represented, and then the token corresponding to the category is added. Based on this, all targets are constructed into a sequence, and the data enhancement algorithm is used to add noise sequences during training. The specific implementation is as follows:
[0131] Step 1, normalize the ground truth in the image to a uniform size.
[0132] Modify the original image according to the aspect ratio of the original image and the set longest side size, such as setting the longest side to 1400; based on the resize ratio of the image, adjust the size of the bou...
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 discloses a target detection model training method, which comprises the following steps: performing feature extraction operation on an image by adopting a backbone network of a to-be-trained model to obtain a multi-scale feature map; encoding the multi-scale feature map based on a multi-scale deformable attention encoding module to obtain encoded image features; performing sequence construction processing based on the correct annotation of the image to obtain a target sequence; decoding the encoded image features and the target sequence based on a multi-scale deformable attention decoding module to obtain a prediction sequence; and performing parameter updating on the to-be-trained model based on a preset loss function, the prediction sequence and the correctly labeled target sequence pair of the image. Therefore, high-complexity feature maps can be processed, and the efficiency and performance of model processing are improved. The invention further discloses a target detection model training device, a server and a computer readable storage medium, which have the above beneficial effects.
Description
technical field [0001] The present application relates to the technical field of machine learning, and in particular to a method for training a target detection model, a training device, a server, and a computer-readable storage medium. Background technique [0002] In related technologies, although PIX2SEQ (a target detection architecture) converts the target detection task into a language model generation task, it still uses a transformer internally. Due to the characteristics of the transformer itself, when processing image features, it will calculate the current position and all other positions. The relationship between the complexity is proportional to the size of the feature map, making it impossible to use large-size or multi-size features, which in turn affects the performance of the model. [0003] Therefore, how to maintain the efficiency in the processing process is a key issue that those skilled in the art pay attention to. Contents of the invention [0004] T...
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 Applications(China)
IPC IPC(8): G06V10/40G06V10/774G06K9/62
Inventor 赵健史宏志金良
Owner INSPUR BEIJING ELECTRONICS INFORMATION IND