| CPC G06T 7/194 (2017.01) [G06V 10/25 (2022.01); G06V 10/44 (2022.01); G06V 10/764 (2022.01); G06V 20/70 (2022.01); G06T 2207/20081 (2013.01)] | 7 Claims |
|
1. An apparatus for generating segmentation models for identifying-objects of interest from images, the apparatus comprising:
an input unit that receives input images;
an object segmentation unit that receives the input images from the input unit, calculates segmentation scores for pixels using the segmentation models, and separates regions of-objects of interest from a background region based on the segmentation scores; and
a learning unit that learns the segmentation models to be used by the object segmentation unit;
a threshold processing unit; and
a stop point determination unit that determines whether to execute training of a next segmentation model in the learning unit,
wherein the learning unit
trains new segmentation models using training data and outputs of existing segmentation models in relation to the training data;
adds the new segmentation models that have been trained to the existing segmentation models, and
repeats the training of the new segmentation models and the adding of the new segmentation models that have been trained to the existing segmentation model,
wherein the threshold processing unit compares the segmentation scores of the pixels in unknown objects of interest detected by the existing segmentation models with a threshold to determine whether the unknown objects of interest are unnecessary objects, and outputs labels of the objects of interest,
wherein the stop point determination unit determines whether to execute training of the next model based on accuracy of estimating a segmentation model trained this time and accuracy of estimating a segmentation model trained last time.
|