Arthropod Image Classification Using User-Feedback Model Training

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing machine learning models for detecting and identifying insects require extensive manual labeling of training data, which is time-consuming and costly, limiting their applicability to specific insect species and necessitating retraining for new species detection.

Innovation Solution

A method involving a first machine learning model for arthropod localization followed by user input for classification, utilizing a second machine learning model to train a classifier based on user feedback, enabling efficient adaptation to new species without extensive manual labeling.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a machine learning model is trained on manually labeled insect images, then the model can detect and identify specific insect species, but the process becomes time-consuming and costly due to extensive manual labeling requirements

Engineering Contradiction:
Improveinsect species identification accuracyVSAvoidmanual labeling time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system enables the model to improve its own classification capabilities through user feedback without requiring extensive manual labeling. Users provide corrections for misclassified insects, and the system automatically retrains the classifier with this feedback, allowing the model to self-improve over time.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system implements a feedback loop where user corrections of misclassified insects are fed back into the training process. The classifier is retrained using user-provided labeled data, continuously improving its accuracy without requiring extensive initial manual labeling of all training data.

Inventive Principle:
Principle #23Feedback

2Measurement precision

If a machine learning model is trained on manually labeled insect images, then the model can identify specific insect species, but the cost of manual labeling increases significantly

Engineering Contradiction:
Improveinsect species identification accuracyVSAvoidmodel training cost
Core Design Contradiction:
Measurement precisionVSEase of manufacture

Solution Approach 1:

The system reduces training costs by enabling the model to learn from user feedback in production rather than requiring expensive pre-labeling of extensive training datasets. The model serves itself by learning from operational data provided by users.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

User feedback on misclassified insects provides free or low-cost training data that continuously improves the model's classification accuracy, replacing the need for expensive manual labeling of large training datasets.

Inventive Principle:
Principle #23Feedback

3Measurement precision

If a machine learning model is trained on specific insect species data, then the model can accurately identify those species, but the model cannot detect and identify other insect species without retraining

Engineering Contradiction:
Improvespecies-specific identification accuracyVSAvoidinsect species range
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system achieves multi-functionality by enabling the same classifier to adapt to multiple insect species through user feedback. The classifier is not limited to pre-trained species but can learn to identify diverse insect species as users provide corrections, making the system universally applicable to various insect types.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The classifier dynamically adapts its capabilities based on user feedback rather than being static after initial training. The system continuously evolves to recognize new insect species as users provide labeled corrections, making the model flexible and adaptable to expanding species ranges.

Inventive Principle:
Principle #15Dynamics

4Measurement precision

If extensive manual labeling is performed for model training, then the model achieves good classification performance, but the productivity of deploying the model decreases

Engineering Contradiction:
Improveclassification performanceVSAvoidmodel deployment speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The model improves its own performance in production through user feedback without requiring extensive pre-labeling. This eliminates the bottleneck of manual labeling and accelerates model deployment, as the system continues to improve autonomously after initial deployment.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary classification using the model's initial training, then uses user feedback on these preliminary results to continuously improve performance. This allows rapid initial deployment followed by progressive improvement, rather than requiring all training to be completed before deployment.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4703918A1Locating and classifying arthropods in images
Publication Date: 2026.03.04 BAYER AG
  • EP4703918A1 patent drawingFigure 1~3(b)
  • EP4703918A1 patent drawingFigure 4(a)~5
  • EP4703918A1 patent drawingFigure 6

AI summary

Systems, methods, and computer programs disclosed herein relate to the localization and classifcation of arthropods in images.