Head model for instance segmentation, instance segmentation model, image segmentation method and device
A technology of head model and segmentation model, which is applied in biological neural network models, computer components, character and pattern recognition, etc., can solve the problems of insufficient segmentation effect of instance segmentation, insufficient accuracy of segmentation information and confidence, etc., and achieve segmentation Accurate information and confidence, fine segmentation effect
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0037] The embodiment of the present application provides a head model for instance segmentation, such as figure 2 As shown, the head model includes a target frame branch 21 and a mask branch 22;
[0038] The target frame branch 21 includes a first branch 211 and a second branch 212, the first branch 211 is used to process the input first feature map T1 to obtain the category information and confidence (class) of the target frame, so The second branch 212 is used to process the first feature map T1 to obtain the position information (box) of the target frame;
[0039] The mask branch 22 is used to process the input second feature map T2 to obtain mask information (M);
[0040] The above-mentioned second feature map T2 is a feature map output by the ROI extraction module, and the first feature map T1 is a feature map obtained by pooling the second feature map T2.
[0041] The location information in this application may be coordinate information, and the confidence degree in...
Embodiment 2
[0069] The present application also provides another head model for instance segmentation, the head model includes a target frame branch, a mask branch and a mask confidence recalculation branch;
[0070] The target frame branch is used to process the input first feature map to obtain category information and confidence of the target frame and position information of the target frame;
[0071] The mask branch is used to process the input second feature map to obtain a third feature map;
[0072] The mask confidence recalculation branch is used to process the second feature map and the fourth feature map to obtain the confidence of the mask branch;
[0073] Wherein, the second feature map is the feature map output by the region of interest ROI extraction module, the first feature map is the feature map obtained after the second feature map is pooled, and the fourth feature map is A feature map obtained after performing a downsampling operation on the third feature map.
[007...
Embodiment 3
[0078] The present application also provides an instance segmentation model, including a main body part, a neck part, a head part and a loss calculation part connected in sequence, and an ROI extraction module is set between the neck part and the head part, The head part adopts any head model in Embodiment 1 and Embodiment 2.
[0079] Wherein, the main part is used to perform convolution calculation on the input image to obtain the first feature map of the image; the neck part is used to process the first feature map to obtain the second feature map; the ROI The extraction module is used to extract the region of interest from the second feature map to obtain a third feature map as the input of the head part.
[0080] For the relevant technical solutions of the embodiments of the present application, please refer to the relevant descriptions in Embodiment 1 and Embodiment 2 and the appended Figures 2 to 4 , to avoid repetition, it will not be repeated here.
[0081] The inst...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More