Ignore Mask Training for Ambiguous Image Feature Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The annotation process for training machine learning models to detect image features like dents is challenging due to the ambiguous nature of dents, leading to inconsistent annotations and contradictory learning patterns for the model.
Innovation Solution
The method involves providing training image data with groundtruth annotations and ignore masks to guide the machine learning model in identifying image features. The ignore masks are generated based on the groundtruth masks to exclude ambiguous pixels from the training process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If random frames are used for annotation without temporal sequence, then the annotation process is simpler and less tedious, but the annotator cannot locate dents accurately and consistency deteriorates
Solution Approach 1:
The system performs preliminary actions by automatically detecting and marking candidate dent regions before the annotation process. This preliminary detection provides spatial context to annotators, enabling them to accurately locate and annotate dents without requiring temporal sequence information, thus resolving the contradiction between ease of operation and measurement precision
2Reliability
If multiple annotators annotate the same frame, then consistency and reliability improve, but the complexity and time required increase
Solution Approach 1:
The system enables self-service by having annotators work independently on pre-processed frames with candidate dent markings. Each annotator can annotate based on their own judgment without requiring coordination with other annotators or temporal context, maintaining reliability while reducing process complexity and time requirements
3Adaptability or versatility
If ambiguous dents are forced to be annotated, then the model learns more patterns, but false positives increase and sensitivity deteriorates
Solution Approach 1:
The system applies local quality by differentiating between high-confidence and low-confidence dent annotations. High-confidence annotations are used for training, while low-confidence ambiguous cases are excluded or handled separately. This selective approach ensures the model learns accurate patterns without being contaminated by false positives from ambiguous detections
Data Source
AI summary
A method of training a machine learning model to identify image features is disclosed including: a. providing training image data, the training image data comprising a plurality of pixels; b. assigning a groundtruth annotation to each pixel, each groundtruth annotation relates to a respective one of the pixels and each groundtruth annotation indicates whether or not that the pixel corresponds with an image feature; c. providing an ignore mask comprising a set of ignore flags; each ignore flag relates to a respective one of the pixels and each ignore flag provides an indication that the pixel should be ignored; d. receiving a prediction value from the machine learning model, each prediction value provides an indication of a probability of the pixel corresponding with an image feature; e. for each pixel which has no ignore flag, determining a loss value based on the prediction value and groundtruth annotation for that pixel.