Object Counting via Partial Region Likelihood and Numerical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image analysis technologies struggle to accurately estimate the number of objects in an image when multiple objects are closely positioned, as they can only indicate information about one object per region, leading to inaccurate counting in congested environments.
Innovation Solution
An estimation apparatus and method that uses a learned model to divide images into partial regions, providing likelihood data and numerical data for each region, enabling the estimation of object numbers by combining likelihood and numerical data to accurately count objects even when they are close to each other.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network outputs a map indicating object position and size for each region, then the model can provide object information for one region, but the number of objects cannot be accurately estimated when multiple objects are positioned in one region
Solution Approach 1:
The image is divided into multiple partial regions, and for each partial region, the model outputs both likelihood data indicating the presence of target objects and numerical data representing the estimated number of objects. This segmentation approach allows the system to handle multiple objects within each region by providing both presence probability and quantity estimation, thereby resolving the limitation of traditional single-object-per-region methods
Solution Approach 2:
The model output structure is changed from traditional single-value predictions to a combination of likelihood data (probability) and numerical data (estimated count). This parameter change enables the system to represent uncertainty and quantity simultaneously, allowing accurate estimation of object numbers even when multiple objects are positioned closely together in the same region
Data Source
AI summary
An estimation apparatus includes an acquisition unit and an estimation unit. The acquisition unit acquires an image. The estimation unit estimates the number of target objects included in a target region being at least part of the acquired image by using a learned model. Input data of the model are an image. Output data of the model include likelihood data and numerical data. The likelihood data indicate a likelihood of a one or more target objects being included in each of a plurality of partial regions acquired by dividing the image. The numerical data indicate an estimated number of target objects for a partial region estimated to include one or more target objects out of the plurality of partial regions. The estimation unit estimates the number of target objects included in a target region by using the likelihood data and the numerical data.