Tattoo Detection Models Using Segmentation for Accuracy Speed Trade-off
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current deep learning-based object detection systems face a trade-off between accuracy and speed, with increasingly deeper architectures improving accuracy but increasing computational expense, and existing methods do not effectively address the challenge of detecting tattoos for identification purposes, especially in real-time applications without high-performance GPUs.
Innovation Solution
The method employs a system that inputs images into a tattoo detection module, selects images with at least three keypoints, manually labels tattoo locations, increases image size for small tattoos, and trains two deep learning models (R-50-FPN and MobileNet-SSD) for both high accuracy and fast execution, using a GPGPU for real-time processing when available.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If increasingly deeper deep learning architectures are employed to improve tattoo detection accuracy, then detection accuracy is improved, but computational expense at runtime increases
Solution Approach 1:
The patent segments the deep learning model into two distinct architectures: a deeper R-50-FPN model for high accuracy detection when computational resources are available, and a lighter MobileNet-SSD model for fast detection when resources are constrained. This segmentation allows the system to select the appropriate model complexity based on runtime requirements, resolving the contradiction between accuracy and computational expense.
Solution Approach 2:
The system dynamically selects between two different deep learning models based on runtime conditions and resource availability. The deeper R-50-FPN architecture is deployed when high accuracy is prioritized and computational resources are sufficient, while the lighter MobileNet-SSD architecture is used when fast execution is required or resources are limited. This dynamic adaptation resolves the static trade-off between accuracy and computational expense.
2Speed
If real-time processing is required for tattoo detection, then processing speed is improved, but detection accuracy may deteriorate
Solution Approach 1:
The patent segments the solution into two specialized models: MobileNet-SSD optimized for real-time processing with lower computational requirements, and R-50-FPN optimized for high accuracy when processing speed is less critical. This segmentation allows the system to match the appropriate model to the specific operational requirement, ensuring both real-time capability and high accuracy are achievable in different contexts.
Solution Approach 2:
The system changes key architectural parameters of the deep learning models to optimize for different priorities. MobileNet-SSD uses lighter parameters and fewer computations for real-time performance, while R-50-FPN uses deeper architectures with more parameters for maximum accuracy. This parameter variation allows the system to achieve both real-time processing and high accuracy depending on the specific use case.
3Productivity
If small tattoo images are used for detection, then data processing efficiency is improved, but detection accuracy deteriorates
Solution Approach 1:
The patent applies preliminary data augmentation and image enhancement techniques to small tattoo images before feeding them into the detection model. This includes padding, resizing, and other preprocessing operations that improve the quality and detectability of small tattoos without significantly increasing processing complexity during inference, thus resolving the contradiction between processing efficiency and detection accuracy for small tattoos.
Data Source
AI summary
A computer-implemented method executed by at least one processor for detecting tattoos on a human body is presented. The method includes inputting a plurality of images into a tattoo detection module, selecting one or more images of the plurality of images including tattoos with at least three keypoints, the at least three keypoints having auxiliary information related to the tattoos, manually labeling tattoo locations in the plurality of images including tattoos to create labeled tattoo images, increasing a size of the labeled tattoo images identified to be below a predetermined threshold by padding a width and height of the labeled tattoo images, training two different tattoo detection deep learning models with the labeled tattoo images defining tattoo training data, and executing either the first tattoo detection deep learning model or the second tattoo detection deep learning model based on a performance of a general-purpose graphical processing unit.