Maximal surface size for a carpentry industry

Application is designed for iOS (iPhone/iPad) platforms and is developed using React Native. Its core functionality revolves around object recognition and geometric analysis to optimize material usage, specifically for scenarios like determining the largest rectangle that can fit within the contours of an irregular object, such as a piece of wood.

Here's how the app operates technically:

1. User Interface: Built with React Native, the application provides a straightforward interface allowing users to either upload an image from their device's storage or capture a new photo directly through the app.
2. Image Processing: Once an image is uploaded or captured, it undergoes processing to identify and analyze the object within. This process leverages the OpenCV library, integrated through Objective-C and C++ methods. The algorithm first identifies the object and its boundaries within the image.
3. Rectangle Optimization: After recognizing the object and its edges, the app calculates the dimensions of the largest possible rectangle that can be inscribed within the object. This feature is particularly useful for material optimization, ensuring minimal waste.
4. Technical Integration: The integration of React Native for the front end and OpenCV for image processing showcases a hybrid approach, utilizing advanced image analysis techniques within a mobile application environment conducive to high-performance and user-friendly interactions.

This application is ideally suited for industries and individuals focused on material conservation and cost-effective resource management, providing a technical solution to enhance material utilization through smart algorithmic analysis.

Subpixel recently took on a challenge to boost efficiency on a production line by enhancing the system for reading QR codes on moving products. Our project focused on making sure the system worked smoothly, even with the existing hardware limitations, and that it could keep up with the fast pace of the production line.

Key Improvements:

• Speed and Accuracy: We made the system faster and more reliable in reading QR codes.
• Handling High Demand: Our solution can manage multiple tasks at once, perfect for keeping things moving quickly.
• Backup Plan for Network Issues: We added a safety net to protect against data loss if the network goes down.
• Working with the Client's System: Integrated seamlessly with the client's database, making everything more streamlined.

Results:

Our work led to a noticeable improvement in how quickly and accurately QR codes are read, keeping production moving without a hitch. Plus, we made sure the system is robust enough to handle network issues without losing important data. The smooth integration with the client's database has made operations even more efficient.

[Note: Images used are for illustration purposes due to confidentiality.]

At Subpixel, we blend smart software solutions with a practical approach, ensuring our projects not only meet the technical needs but are also user-friendly and reliable.

Layer counting

Subpixel embarked on a challenging project to automate the counting of stacked leather sheets, a critical component in automotive part production. Leveraging deep learning techniques, our team developed a sophisticated solution designed to address the unique complexities of this task.

Project Challenges:

• Variable Layer Thickness: The leather sheets varied significantly in thickness, presenting a considerable challenge for accurate counting.
• Detection of Layer Borders: Identifying the borders between layers proved to be exceptionally difficult due to the subtle distinctions in texture and color.

Innovative Approach:

- Deep Learning Techniques: We applied advanced deep learning algorithms tailored to recognize and differentiate the nuanced characteristics of each leather sheet, despite the variations in thickness and the elusive nature of the layer borders.

Results:

The solution enhanced the efficiency and accuracy of the counting process for stacked leather sheets, crucial for maintaining inventory accuracy and production scheduling in automotive manufacturing. This project not only demonstrated Subpixel's capability to solve highly specific and complex problems through innovative technology but also improved a key manufacturing process for our client.

[Note: Imagery is representative due to a non-disclosure agreement.]

Project Overview:

Subpixel successfully spearheaded a project that stands out in the field of AI communication, emphasizing the precision of prompt engineering to generate high-quality, reliable AI responses. Our goal was to mitigate the occurrence of 'hallucinations' or inaccurate information, often a challenge in language model outputs.

Solutions and Technical Excellence:

• Prompt Engineering: Developed advanced techniques for crafting prompts that guide AI to deliver accurate and contextually relevant responses, ensuring information reliability.
• MongoDB Vector Databases: Implemented a robust MongoDB vector database architecture to create a private knowledge base, enhancing the AI’s reference material for improved response quality.
• Communication Source Integration: Integrated a system to extract and incorporate knowledge from various communication mediums such as emails and chat systems, enriching the AI's understanding and response accuracy.
• Source Display for LLM Answers: Introduced a feature that traces the lineage of AI responses, providing users with the source documents related to their queries, thereby increasing transparency and trust in AI-generated information.

Client Benefits:

The client now benefits from an AI system with enhanced accuracy, capable of tapping into a rich vein of organized knowledge, and delivering responses that are not only relevant but also verifiable. This has fostered a greater level of confidence in AI interactions within their operations.

Why Subpixel:

At Subpixel, we excel in advancing technology to enhance AI communications' reliability. Our prompt engineering and data management expertise streamline AI integration for businesses, ensuring seamless and trustworthy operations.

Ready for AI You Can Trust?

Discover how Subpixel's innovative solutions can elevate your AI capabilities and provide you with insights you can rely on.

Development of CAD tool

Revolutionizing 3D CAD Model Manipulation:

Our team at Subpixel was tasked by a client to enhance their 3D CAD model interaction. We developed an advanced offline tool, addressing the need for rapid loading and sophisticated manipulation of 3D models in industry, crucial for efficient processes from product design to production. 

Solutions:

• Script for Format Conversion: Ensured seamless model compatibility.
• Parallel Rendering: Significantly improved loading and rendering speed, enhancing user experience.
• Mesh Creation: Transformed individual polygons into detailed meshes for richer model fidelity.
• Advanced Manipulation Tools: Introduced cutting-edge features including precise model cutting, internal element calculation, laser scanning simulation, and dynamic manipulation of polygons, meshes, and OpenGL scenes. We also enabled normal adjustment and merged multiple 3D formats into a singular, optimized model.

Outcomes:

These innovations led to a transformative improvement in the client's design process, marked by faster operations, superior model quality, and unprecedented control over model analysis. Our project set new industry standards for 3D modeling tools, showcasing Subpixel's commitment to technological excellence and client success.

Transform Your Product Design:

Contact us to explore how our solutions can elevate your 3D modeling capabilities.

Subpixel has delivered a specialized defect detection and classification solution for small electronic components, enhancing the quality control process in electronics manufacturing. Designed to inspect components between 4 to 13 mm, our system adeptly identifies a range of defects, including chip-offs, discoloration, scratches, cracks, and spots.

Project Insights:

• Comprehensive Defect Detection: Our technology is engineered to accurately detect a variety of imperfections, crucial for maintaining high quality in production.
• Precision Measurement: The system ensures precise measurements, vital for consistent product quality.
• Use of Telecentric Lenses: Telecentric lenses are employed to maintain image consistency across the view, essential for reliable defect identification.
• Enhanced Imaging Resolution: With a detail resolution of 10 µm per pixel, the system excels in spotting very small defects that standard inspections might overlook.
•  

Outcomes:

The deployment of our system has made a noticeable difference in the client's manufacturing process, offering a method to better identify and classify defects, thus aiming to reduce waste and improve the quality of the output. The incorporation of telecentric lenses and high-resolution imaging marks a step forward in quality assurance practices.

[Note: Due to a non-disclosure agreement, the imagery is illustrative.]

This project underscores Subpixel's role in advancing quality control technologies in the electronics manufacturing sector, striking a balance between innovation and practicality.

High-precision measurement

Subpixel took on the challenge of overhauling an established production machine with our bespoke computer vision software, aimed at high precision measurement of electronic components. Our client required exacting standards where component sizes ranged from 4 to 13mm, tolerances were set to a stringent 200 microns, and deviations were to be kept under 1 micron. 

Our solution not only met these specifications but included a robust defect detection system, further ensuring quality control in the production line. The desktop application, crafted with C# .NET and leveraging the powerful MVTec Halcon library, stands as a testament to our commitment to precision and reliability.

Due to confidentiality, we use symbolic imagery to represent our work's impact. This project underscores Subpixel's capability to modernize and enhance production processes, ensuring our clients maintain the cutting edge in manufacturing excellence.

[Project visual is symbolic due to NDA.]

Project Overview:

This project leverages advanced computer vision techniques and classification algorithms to extract rich product data and pricing insights from physical or digital catalogs.

Solutions and Technical Excellence:

• Image Acquisition and Preprocessing: High-resolution digital scans or photographs of catalog pages are captured and preprocessed to ensure optimal image quality for analysis.
• Object Detection and Recognition: A deep learning model trained on a massive dataset of labeled catalog images is used to identify and localize specific objects within each catalog page. This includes products, branding elements, and price tags.
• Optical Character Recognition (OCR): Extracts text data from the catalog, such as product descriptions, brand names, and prices.
• Price Analysis and Comparison: Extracted pricing information is normalized and formatted to enable comparison across different products and brands within the catalog, and potentially with external pricing data sources.
• Trend Identification: Historical pricing data from previous catalogs can be incorporated to identify price trends for specific products or categories over time.

Client Benefits:

• Automated Data Extraction: Manual data entry from catalogs is eliminated, saving significant time and resources.
• Improved Data Accuracy: Eliminates the risk of human error in data entry.
• Enhanced Product Insights: Extracted data can be used for product assortment optimization, pricing strategies, and competitor analysis.
• Trend Monitoring: Enables proactive identification of price changes and trends.

Why Choose This Solution?

This AI-powered solution offers a fast, accurate, and scalable approach to extracting valuable product and pricing data from catalogs. It empowers businesses to optimize their pricing strategies, gain a deeper understanding of their product assortment, and make data-driven decisions to maximize profitability.