AWS
MarkNow
AI-Based Trademark Recognition & Image Comparison Model Development
This project was undertaken to develop an automated trademark detection and similarity analysis platform that integrates cutting-edge Computer Vision and deep learning technologies to revolutionize the efficiency of intellectual property management.
MarkNow goes beyond simple image matching: the AI model converts input trademarks (logos/symbols) into high-dimensional vector feature embedding data and quantitatively calculates similarity against existing registered trademarks.
Technically, YOLO-based object detection algorithms are precisely combined with CNN-based feature extraction networks to achieve accurate trademark identification even within complex images.
Advanced data augmentation and preprocessing techniques were applied to handle various environmental variables, maintaining recognition accuracy above 90% even under low-resolution or poor lighting conditions.
Through this system, TOSKY completed an intelligent IP solution that enables early identification of duplicate and similar trademark issues during new trademark registration and dramatically improves the efficiency of legal and administrative review processes.
Key Features
01
YOLO/CNN-Based Precision Object Detection
Accurately identifies trademark locations within images and extracts unique feature points through state-of-the-art deep learning networks
02
Vector Embedding Similarity Measurement
Converts trademarks into numerical data and derives objective, quantitative similarity scores against existing databases
03
High-Reliability Environment-Adaptive AI Model
Ensures stable accuracy above 90% even under challenging conditions such as angle and resolution variations through data augmentation