Image Region Identification Using Neural Network Feature Comparison

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current technologies lack an efficient method to identify and retrieve images that depict specific items, such as clothing or other personal items, from a collection of images, making it difficult for users to recall past outfits or find matching items.

Innovation Solution

The use of trained convolutional neural networks and Siamese networks to compare features of items with regions-of-interest (ROIs) in images, determining similarity and outputting images that match the query item, with adjustable ROI sizes based on item size and image depth.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If traditional image search methods are used, then users can search for images, but they cannot efficiently identify and retrieve images of specific items from a collection

Engineering Contradiction:
Improveimage identification efficiencyVSAvoiditem retrieval capability
Core Design Contradiction:
ProductivityVSEase of operation

Solution Approach 1:

The patent divides the image into multiple regions of interest (ROIs) corresponding to different body parts (head, torso, legs, etc.), allowing the system to selectively search for items in specific regions rather than analyzing the entire image, thereby improving identification efficiency and accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces a region-of-interest (ROI) as an intermediary between the user's item query and the image data, enabling the system to map user-friendly item descriptions to specific image regions and retrieve relevant images efficiently

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If the system analyzes the entire image for item identification, then it can find all possible matches, but it increases processing time and computational complexity

Engineering Contradiction:
Improveitem detection accuracyVSAvoidimage processing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent segments the image into multiple ROIs representing different body parts, allowing the system to process only the relevant regions where items are likely to appear, significantly reducing processing time while maintaining detection accuracy through targeted analysis

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies partial action by analyzing only the necessary portions of the image (ROIs) rather than the entire image, optimizing the balance between processing thoroughness and computational efficiency

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS9953242B1Identifying items in images using regions-of-interest
Publication Date: 2018.04.24 AMAZON TECH INC
  • US9953242B1 patent drawing
  • US9953242B1 patent drawing
  • US9953242B1 patent drawing

AI summary

The techniques described herein may identify images that likely depict one or more items by comparing features of the items to features of different regions-of-interest (ROIs) of the images. When a user requests to identify images that depict a particular item, the techniques may determine a region-of-interest (ROI) size based on the size of the requested item. The techniques may then search multiple images using the ROI size.