Innovative Tool for 3D Model Customization

With recent progress in Generative AI, it's now easier to change 3D models automatically. However, most current methods make changes to the entire model, which might affect how the model works when made in real life. For example, if you change the base of a vase, it could tip over and not work as intended. We are introducing a new method to separate 3D models into parts that are functional and parts that are just for looks. This way, users can modify only the aesthetic parts without changing how the model functions.

To create this method, we first looked at 1,000 different models from a popular 3D printing site called Thingiverse. Based on this analysis, we developed a way to automatically categorize the parts of 3D models. Our system, called Style2Fab, allows users to style their 3D models while making sure they still work as intended.

#The Problem

A common issue for many creators is how to change or personalize designs that they find online. While there are many ready-to-print models available, users usually have limited options to customize these designs. Recent improvements in AI allow for these kinds of changes, but users still face challenges. They must identify which parts of a model are important for its function and which parts are just for looks. This can be a very hard task, especially for those unfamiliar with the designs.

Some tools allow users to label functions in CAD software, but many models shared online do not come with this necessary information. To make it easier for users, we are presenting a new way to automatically break down 3D models into functional and aesthetic parts. Our method gives creators the ability to style their 3D models while keeping the original function intact.

#Taxonomy of Functionality

We developed a classification method for 3D models based on our study of designs on Thingiverse. We identified three main categories for the parts of a model:

1. Aesthetic: These parts are just for looks and do not affect how the model works.
2. Internally Functional: These parts relate to how different components fit together, like the pieces of a puzzle.
3. Externally Functional: These parts interact with the outside world and are crucial for the model's function.

Using this classification, we created a method to automatically identify these segments in 3D models.

#The Style2Fab System

Style2Fab is an interactive tool that enables users to manipulate 3D designs without changing how they work. The system first segments the model into functional and aesthetic parts. Users can then apply their desired styles only to the aesthetic areas. The system uses advanced techniques to ensure that the model's functionality is preserved.

For example, a user may want to change the appearance of a self-watering planter. They recognize that the base needs to stay flat for stability, along with other functional parts. Using Style2Fab, the system labels these important parts for the user and allows them to apply styles only to the outer sections, ensuring the planter will still work.

#Supporting Creators in 3D Printing

When 3D printing became popular, many people began sharing their designs online. Platforms like Thingiverse became great places for novices to learn and explore. However, many creators struggle to make changes because available tools often require advanced skills.

Several tools have been developed to assist creators. For example, some tools automate the recombination of elements in 3D printed designs or aid in creating connectors for different objects. Other systems help integrate real-world objects into existing 3D designs.

Despite these developments, most tools do not focus on functionality. Our approach aims to support users by creating clear classifications in 3D designs, allowing for easier personalization.

#Functionality-Aware Design in 3D Models

Making modifications to 3D models requires consideration of the specific functions intended for those designs. Some have suggested methods to assist in this process. For example, one tool, called Meshmixer, allows for direct manipulation of models meant for 3D printing.

Many existing models lack the necessary data for effective customization. To help address this issue, our system segments models into aesthetic and functional parts, which makes it easier for users to modify the looks without affecting how the model performs.

#Method for Segmenting Functionality

Our method focuses on breaking down the 3D models into meaningful segments. We look at the model's geometry and analyze how the different parts fit together. By doing so, we can identify which segments are important for function and which are purely decorative.

We first segment the models based on how they are designed. The Segmentation process involves examining the connections and features of each mesh.

#Data Collection

To create our method, we collected a dataset comprising the most popular designs on Thingiverse. We analyzed 1,000 different models to gain a better understanding of their Functionalities. Most models were shared in formats that make them difficult to edit, so we focused on 3D meshes that fit our criteria.

After cleaning up the dataset, we were left with 993 unique models divided into various categories. This allowed us to draw connections and categorize Aesthetics and functions across the range of models we studied.

#Developing Our Taxonomy

Using the collected data, we performed an iterative coding process to classify the functionalities of the designs. We found two major categories:

1. Artifacts: These are mainly aesthetic items with little to no functional parts.
2. Task-Related Models: These are designed specifically for tasks, such as a stand or holder.

We further divided the models into four basic categories based on the internal and external aspects of functionality.

#Segmenting and Classifying 3D Models

Our method for segmenting models focuses on isolating segments that serve specific functions. This allows users to modify the visual aspects of the design without affecting essential pieces. Our approach leverages both external and internal contexts to classify the various parts of a model effectively.

We developed a systematic process to classify segments as either functional or aesthetic based on their geometry and connections to other segments.

#Evaluating Our Method

To assess the effectiveness of our segmentation and classification method, we created Style2Fab, which integrates this functionality into a user-friendly interface. We conducted user studies to measure how well our system supports users in separating and styling functional elements of their models.

In the studies, participants were tasked with identifying segments in both automated and manual conditions. The results showed that users were faster and more accurate when using the Style2Fab system.

#User Experience with Style2Fab

The overall feedback from users was positive. Many found the tool helpful in identifying functional segments and appreciated the automation support. Participants reported that using Style2Fab made the process of modifying models more straightforward and efficient.

#Application Scenarios

We present several real-world examples to demonstrate how Style2Fab can be used in various contexts.

#Home Decor

One example is a self-watering planter that users can style while keeping the function intact. Another example is a drink dispenser that allows for aesthetic changes without altering how it distributes liquid.

#Medical Applications

Style2Fab is also beneficial in the medical field, where personalization is essential. For instance, a thumb splint can be styled to be more visually appealing while maintaining its essential support function.

#Personalized Accessories

The tool can also be used to create unique accessories, such as custom cases for electronic devices. Users can enhance the designs while ensuring that they remain practical and functional.

#Limitations and Future Directions

While our functionality-aware segmentation method is effective, there are areas for improvement. For instance, our classification system currently relies on specific topological features, which may not capture all aspects of functionality.

An expanded definition of functionality could improve the method. Additionally, creating larger datasets with diverse models would enhance the robustness and accuracy of our system.

#Conclusion

In summary, we introduced a new system that allows creators to modify and style 3D models while keeping their function intact. This method focuses on understanding the different segments in a model and how they relate to functionality. Through user studies and real-world examples, we showed that our approach is beneficial for both novice and experienced makers.

Ultimately, Style2Fab aims to empower users in their creative journeys, making it easier for everyone to personalize their designs. As 3D printing technology continues to evolve, we believe our approach will play a valuable role in shaping how creators interact with digital designs.