Mobile Object Navigation Using Confidence-Ranked Object Dialogue
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection systems in mobile objects face recognition errors and misunderstandings due to varying degrees of certainty, leading to increased dialogue turns and reduced user convenience.
Innovation Solution
A mobile object control device that utilizes a learned model to output objects with corresponding degrees of certainty, sequentially selects objects based on score values considering certainty, horizontal position, classification, and moving tendency, and makes inquiries to the user to reduce dialogue turns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional object detection is used to detect objects in the input image, then the system can identify objects, but recognition errors occur and misunderstandings happen between user and system
Solution Approach 1:
The system presents detected objects with their certainty scores to the user and receives feedback through dialogue interactions. The user confirms or corrects the system's object identification, allowing the system to learn from feedback and improve future detections, thereby resolving recognition errors and misunderstandings.
Solution Approach 2:
The system performs preliminary object detection and sorts candidates by certainty score before presenting them to the user. This preliminary sorting action ensures that the most likely correct objects are presented first, reducing the chance of misunderstanding and improving dialogue efficiency.
2Measurement precision
If the system asks multiple questions to confirm the object, then the object can be accurately specified, but the number of utterance turns increases and user convenience decreases
Solution Approach 1:
The system performs preliminary sorting of detected objects by certainty score before initiating dialogue. By pre-organizing candidates in descending order of confidence, the system minimizes the number of dialogue turns needed, as the user is presented with the most likely correct object first, reducing dialogue duration while maintaining accuracy.
Solution Approach 2:
The system changes the parameter of object presentation by sorting based on certainty scores rather than arbitrary or sequential ordering. This parameter change optimizes the dialogue process by ensuring that the most confident detections are verified first, reducing the average number of utterance turns required.
3Adaptability or versatility
If the system detects all possible objects, then comprehensive object identification is achieved, but the complexity of selecting the correct object increases
Solution Approach 1:
The system segments the set of detected objects by sorting them into a hierarchical order based on certainty scores. This segmentation transforms a complex set of all possible objects into a structured sequence where the most likely correct objects are at the top, simplifying the selection process while maintaining comprehensive detection coverage.
Solution Approach 2:
The system performs preliminary sorting and filtering of detected objects before presenting them to the user. By pre-organizing objects based on detection confidence, the system reduces the complexity of object selection while maintaining adaptability to detect various types of objects in the input image.
Data Source
AI summary
A mobile object control device acquires a captured image obtained by capturing an image of surroundings of the mobile object by a camera mounted on a mobile object and an input directive sentence input by a user of the mobile object, inputs, when an image and a directive sentence are input, the captured image and the input directive sentence into a learned model learned to output one or more objects corresponding to the directive sentence in the image together with corresponding degrees of certainty to detect the one or more objects and the corresponding degrees of certainty, sequentially selects the one or more objects based on at least the degree of certainty and makes an inquiry to a user of the mobile object, and causes the mobile object to travel to an indicated position in the input directive sentence, which is specified based on a result of the inquiry.