Method and system for image segmentation and identification

Inventors

Peng, Yu

Assignees

Curvebeam AI Ltd

Abstract

An image segmentation method system, the system comprising: a training subsystem configured to train a segmentation machine learning model using annotated training data comprising images associated with respective segmentation annotations, so as to generate a trained segmentation machine learning model; a model evaluator; and a segmentation subsystem configured to perform segmentation of a structure or material in an image using the trained segmentation machine learning model. The model evaluator is configured to evaluate the segmentation machine learning model by (i) controlling the segmentation subsystem to segment at least one evaluation image associated with an existing segmentation annotation using the segmentation machine learning model and thereby generate a segmentation of the annotated evaluation image, and (ii) forming a comparison of the segmentation of the annotated evaluation image and the existing segmentation annotation. The method includes deploying the trained segmentation machine learning model for use if the comparison indicates that the segmentation machine learning model is satisfactory.

Core Innovation

The invention provides an image annotation system and an image annotation method that receive at least one unannotated or partially annotated image and form an annotated training image. The annotator generates one or more candidate image annotations for the image and receives inputs identifying one or more portions of the candidate annotations as being satisfactory, and the annotated training image is formed from at least the identified portions.

Candidate image annotations are selected from one or more generated candidate image annotations using user inputs that identify satisfactory portions, enabling an annotated training image to be constructed from only the satisfactory parts. In certain implementations, the candidate image annotations are generated using non-machine learning image processing, and in other implementations the candidate image annotations are generated using a segmentation machine learning model.

The disclosed system further integrates forming annotated training images with training and evaluation of a segmentation machine learning model. A trained segmentation machine learning model segments an annotated evaluation image, compares the segmented result with one or more existing segmentation annotations, optionally within a predefined threshold, and deploys or releases the model when the comparison indicates the model is satisfactory, with continuing training by adding annotated training data.

Claims Coverage

The document includes two independent claims. Across these claims, the inventive concepts focus on forming annotated training images from candidate annotations using user inputs identifying satisfactory portions, with optional dependence on how candidate annotations are generated and on an end-to-end segmentation training, evaluation, and deployment workflow.

Overall claim coverage centers on using candidate image annotations plus user-provided identification of satisfactory portions to construct annotated training images, with further coverage of candidate generation using non-machine learning image processing or a segmentation machine learning model, and an evaluation and comparison-driven training and deployment workflow for a trained segmentation machine learning model.

Stated Advantages

Documented Applications