Skip to main content
All projectsDelivered

AI-powered Vision Metrics

We at Saiwa Inc. are thrilled to announce the successful completion of an innovative project focused on the measurement of interpupillary distance (IPD) and pupil height. This groundbreaking solution was developed in collaboration with Innovation Venture Farm to address the growing need for precision in optical device fitting and design. By leveraging cutting-edge AI techniques and the Mediapipe framework, our solution detects and calculates critical parameters such as the center of the pupil and the lowest point of glasses lenses. These measurements are crucial for applications in eyeglasses, VR/AR headsets, and other optical technologies. The results are delivered with unmatched accuracy and efficiency, empowering manufacturers to deliver tailored products to their users. This project also integrates Saiwa’s Annotation Services, which ensure data labeling precision and scalability, making it easier for users to achieve reliable outcomes.

Ordered byAI-innovate
AI-Powered IPD and Pupil Height Measurement
Timeline
Mar 2024 – May 2024
Solution
Platform
Country

Measurement Method of Interpupillary Distance and Pupil Height

Overview

Saiwa introduces an innovative approach to measuring interpupillary distance (IPD) and pupil height, leveraging state-of-the-art AI technologies. These measurements are critical for ensuring the comfort and effectiveness of optical devices such as eyeglasses, virtual reality (VR) headsets, and augmented reality (AR) wearables. Saiwa’s solution redefines precision in this domain, offering a reliable method that can revolutionize how optical measurements are performed globally.

Our goal is to empower manufacturers and designers with tools that optimize product fit, enhance user experience, and meet the increasing demands of modern optical technology. By combining cutting-edge algorithms with a robust dataset, Saiwa ensures unparalleled accuracy and scalability.

glasses_temp.jpg
Image Source: Freepik.com

Challenges and Our Solution

Accurate IPD and pupil height measurements have traditionally been complex and prone to error, especially when reliant on manual methods or inconsistent data. Saiwa’s solution addresses these challenges by:

  • Using Advanced Deep Learning Models: Incorporating the Mask R-CNN framework and BlendMask algorithm, our method ensures precise detection of key features in facial images, such as eyeglasses and pupil centers.

  • Utilizing Custom-Labeled Data: Through Saiwa Annotation Services, every image in the dataset is meticulously labeled, capturing the nuances needed for precise calculations.

  • Enabling Seamless Scalability: Designed with adaptability in mind, this method supports diverse datasets and hardware setups, making it suitable for a wide range of applications.

glasses_temp2.jpg
Image Source: Freepik.com

Data Preparation and Tools

Saiwa’s approach begins with a meticulously curated dataset derived from a Kaggle competition. The dataset features images of synthetic individuals generated by a Generative Adversarial Neural Network (GAN), ensuring diversity and coverage. Key steps include:

  • Data Segmentation: Images containing eyeglasses were isolated and segmented to focus on relevant features.

  • Annotation: Saiwa’s proprietary annotation tools were employed to label images with high precision, ensuring the coordinates of critical features, such as the lowest point of the glasses lenses, were accurately captured.

face-1636_im1_dis.png
Image Source: kaggle.com

Methodology

Saiwa’s workflow integrates AI-powered algorithms and user-centric processes to ensure efficiency and reliability. The main steps include:

  1. Detection of Glasses Lenses: The BlendMask deep learning algorithm identifies the lenses and marks their lowest points (“x_low”, “y_low”).

  2. Identification of Pupil Centers: Mediapipe’s advanced AI-powered detection precisely locates the center of the pupil (“x_c”, “y_c”).

  3. Pupil Height Calculation: Pupil height (PH) is computed as the vertical distance between the pupil center and the lowest lens point:

  4. Pixel-to-Millimeter Conversion: Measurements are standardized by using a reference object at a fixed camera distance of 30 cm, ensuring reliable real-world scaling.

with-glasses.jpg
Image Source: Freepik.com

Key Results

Saiwa’s measurement method has demonstrated:

glsss.png
Images Source: kaggle.com

Trusted by Our Customers

Innovation Venture Farm has already adopted Saiwa’s solution for their advanced optical measurement needs, showcasing its versatility and reliability in real-world applications.

  • Unmatched Precision: High accuracy in detecting and measuring features essential for IPD and pupil height calculations.

  • Improved Efficiency: A streamlined process that minimizes manual intervention and reduces errors.

  • Versatility: Applications extend beyond eyeglasses to VR/AR headsets, industrial goggles, and medical diagnostic tools.

Why Choose Saiwa’s Solution?

Saiwa’s expertise in leveraging AI for optical measurements offers clients a competitive edge:

  • Precision That Matters: Accurate measurements directly translate to enhanced user satisfaction and reduced returns for ill-fitting optical devices.

  • Scalability for Growth: Whether you’re a boutique eyewear brand or a global AR/VR provider, Saiwa’s solution scales to your needs.

  • Future-Ready Innovation: As technology evolves, Saiwa’s commitment to integrating advanced AI models like BlendMask and Cascade Mask R-CNN ensures you stay ahead of the curve.

Looking Ahead

Saiwa continues to push boundaries in optical measurement by:

  • Exploring New Algorithms: Transitioning to advanced models for even greater accuracy and efficiency.

  • Enhancing Datasets: Expanding training data to improve robustness and reliability across diverse scenarios.

  • Refining Parameters: Optimizing learning rates, batch sizes, and other model configurations for peak performance.

Get Started

Ready to revolutionize your optical measurement processes? Discover how Saiwa can elevate your product design and user experience. Visit our projects page to learn more and join the future of optical innovation.

Share this project

Help others discover this work

Related Projects

  • Paper Surface Defect Detection
    AI-innovate

    Ordered by

    AI-innovate

    Paper Surface Defect Detection

    May 2023→Aug 2023
    Discover how Saiwa Inc. leveraged cutting-edge AI and deep learning to detect surface defects on both white and black paper with high precision. This project revolutionized quality control for a leading paper manufacturer—improving accuracy, reducing waste, and enabling real-time defect insights.
  • Aluminum Surface Defect Detection
    We at Saiwa Inc. are pleased to announce that we have successfully completed an aluminium surface defect detection project in collaboration with AI-innovate Company for CastTechnology in Canada. In this project, using machine learning techniques and the networks we provide in our Anomaly Detection service, we detect and localize the location of micro and macro defects on a casting line, including: crack, frost, frost patch, longitude frost and mold oscillation. This service is delivered via a simple user interface where users can run the defect detection APIs.
  • Corrective Exercise mobile application
    This is a corrective exercise mobile application which is developed for Android platform. Corrective Exercise is a technique that leverages an understanding of anatomy, kinesiology, and biomechanics to address and fix movement compensations and imbalances to improve the overall quality of movement during workouts and in everyday life. This technique is used to help assess and determine the root cause of imbalances and faulty movement patterns that lead to issues with posture, balance, and total body coordination.
  • Surface Defect Detection
    AI Innovate

    Ordered by

    AI Innovate

    Surface Defect Detection

    Sep 2021→Jan 2022
    For a long time, product quality control relies on manual examination in the manufacturing field. In the past several decades, automated surface inspection (ASI) have become more widespread as hardware and software technologies have developed rapidly. To lower labor costs and enhance examination efficiency, more factories start to employ embedded machines for product inspection. ASI platforms use special light sources and industrial cameras to capture the images of the product surface, and machine/computer vision technologies to filter out defective products, which can reduce labor greatly. High-performance cross-product ASI algorithms are urgently needed in manufacturing.
  • RNA-cleaving Fluorescent DNAzymes (RFDs)
    RFDs are DNA sequences that induce a cleavage (cutting) reaction in a substrate strand in response to the presence of a target. In this project, we at saiwa team segmented and measured regions of interest (the dots that constitute the microarray) relative to the background in printed RFD images. Printed microarrays can be used to test for the presence of specific targets (i.e., bacteria). We used around 100 RFD images that were provided by Didar Lab., McMaster uni.​
  • Drone-Based Cotton Stand Detection, Counting, and Mapping for Precision Agriculture
    Discover how Saiwa developed an AI-powered solution that detects and counts individual cotton plants from drone imagery, maps their locations, and generates John Deere Operations Center-compatible outputs to support automated weed management in cotton fields.