Title: VIIM - a digital fashion community service powered by generative AI and blockchain
Author: Hyekyung Ko / discord id : hko8103
Date posted: 2023/07/19
- VICEVERSA is developing a digital fashion community platform called VIIM, leveraging generative AI and blockchain technology.
- Fashion brands have the opportunity to create and train their own AI models, which can then be secured and monetized via blockchain technology.
- The platform allows users to design and share their own digital fashion content using the generative AI provided by both VIIM and fashion brands.
- The GPU resources requested will be primarily used for training generative AI models catered to specific fashion brands.
- The contribution from the AI Network in terms of GPU resources would significantly enhance the development and sophistication of generative AI within the fashion industry.
- 1,000 coupons will be offered for using ‘AI Photoshoot’ service in our platform, which is a paid-service. Through our ‘AI Photoshoot’ service, users can get images featuring themselves as models for a variety of fashion styles. We will provide this service within our ‘VIIM’ mobile application.
- We plan to highlight AI Network as our provider of GPU resources during the promotional and marketing activities for our service.
- Our contribution to the AI network ecosystem will be evaluated, considering resource provision from ecosystem participants.
- Description: 3 GPU server for training
- Resource type: A100 GPU with 1TB of storage.
- Amount: 3
- Purpose of use : AI model training
VICEVERSA is a team of AI, blockchain, and fashion professionals. Our mission is to empower consumers to create the products they wish to consume, and enhance their engagement with fashion brands using AI and blockchain technologies. Our focus on fashion stems from the understanding that fashion is one of the most frequently consumed items and a critical aspect of our daily lives. We believe that AI and blockchain technology have the potential to revolutionize the fashion industry, given its vast scale and the widespread influence it exerts.
The modern consumer in the fashion industry is evolving from a mere buyer of clothes to an active participant in expressing their lifestyle and identity through the clothes they own. To cater to these evolving consumers, fashion brands need to reflect their preferences and values. Fans of fashion brands desire active participation and engagement with the brands they love, emphasizing the importance of community building for these brands. Generative AI serves as a useful tool for fans to become an integral part of the fashion brands they admire, enabling them to create designs that reflect their preferences.
However, generative AI is currently insufficient in generating detailed fashion styles or products. Our solution to this challenge is the development and training of our own AI models tailored to the fashion industry and individual fashion brands. We anticipate that fashion enthusiasts will find great value in our services as we develop specialized AI models for fashion.
As we continue to develop and train our AI models, the computational resources required for training and deploying these models have increased significantly. Access to AI Network’s powerful GPUs is essential for us to maintain our competitive edge in the rapidly evolving AI landscape. Using AI Network’s GPU resources, we aim to develop generative AI models for fashion that will enable the creation of high-quality digital fashion content.
- The project’s goal is to develop and refine generative AI models specifically tailored for the fashion industry and individual fashion brands.
- The output of these models should effectively represent detailed and accurate fashion styles or products.
- Text2Fashion: We aim to develop AI models trained on fashion datasets that our team will collect and organize. The trained AI models are designed to describe and generate fashion styles and products with more depth and detail.
- Text2Brand: We will train AI models for each specific fashion brand. These models will generate brand-specific styles. For example, a GUCCI AI model will generate clothes and products in the unique style of GUCCI, even those that don’t exist in reality.
- We will experiment with sample datasets and conduct research to minimize development time.
- We will collect and preprocess a diverse fashion dataset that can be used for training the AI models.
- We will train the deep learning models on the preprocessed data.
- We will apply knowledge distillation techniques to create compact versions of the model while maintaining high performance levels.
Week 1-2: Prototyping
- Define project requirements and objectives.
- Review existing research on training AI models and methods for effective and efficient training.
- Develop initial fashion datasets, with a focus on expressing details such as fit, textile, and decoration.
Week 3-5: Data Collection and Preprocessing
- Identify and curate relevant datasets for training and validating the generative AI models.
- Collect additional data through partnerships.
- Ensure data diversity and representation to avoid biases and improve the robustness of the AI model.
- Clean and preprocess the collected data, including categories such as fit, textile, decoration, and other details.
- Split the data into training, validation, and testing sets to facilitate model evaluation and performance monitoring.
Week 6-15: Model Training
- Develop and train generative deep learning models using the preprocessed data and secured GPU resources.
- Experiment with various model architectures, hyperparameters, and training strategies to optimize performance and minimize overfitting.
- Monitor training progress and make adjustments as necessary to ensure convergence and stability of the generative models.
Week 15-20: Model Validation & Iteration
- Evaluate the performance of the trained generative models on the validation and testing datasets using both quantitative and qualitative assessments.
- Identify areas for improvement and iterate on the models, addressing any issues with the realism, diversity, or naturalness of the generated fashion style expressions.
- Conduct pilot tests and demonstrations with users and fashion professionals to gather feedback on the usability, effectiveness, and expressiveness of the AI models.
Upon completion of these milestones, we will continue refining and expanding our AI models based on the feedback and insights gained during the validation phase. Additionally, we will explore potential partnerships, licensing opportunities, and integration of our technology into existing platforms and applications.
Our project aims to transform the way we interact with fashion and brands. We leverage the power of artificial intelligence and machine learning to create realistic and expressive fashion styles and products. Our platform offers AI models for general fashion and for each individual fashion brand, working in tandem to provide seamless and immersive experiences for users across the fashion industry.
Text2Fashion: Our AI models for fashion are deep learning models that analyze fashion styles and products to generate digital fashion content. They visualize the unique sartorial dreams and styles of fashion enthusiasts. These enthusiasts can thus become part of fashion brands by expressing their creative ideas for fashion. Additionally, these models could serve as sources of inspiration for fashion designers, enabling them to design beyond their current imagination.
Text2Brand: We empower fashion brands to create their own AI models, trained with their specific collections and content. Brands can encourage their fans to create clothes that reflect their identity, fostering inspiration and a closer connection with their fanbase. This represents a novel method of communication and community building with consumers.
Our platform provides an immersive and engaging experience for fashion lovers across various brands. Through our technology, we envision a future where digital fashion communities are more vibrant, expressive, and impactful than ever before.
- Generation of High-Quality Digital Fashion Content: Our models aim to generate visually accurate and detailed clothing.
- Accurate Interpretation of Input Text: AI models must generate contextually appropriate styles and products that align with the user’s intent.
- User Satisfaction and Engagement: The project’s success hinges on user satisfaction with the quality and accuracy of AI-generated fashion styles and products.
- Expression of significant styles from fashion brands: Brands’ AI models should generate fashion content reflecting the styles of each brand.
- Scalability and Adaptability: The AI models should be scalable to accommodate a growing user base and adaptable to incorporate new features and improvements based on user feedback and advancements in AI research.
Myunghoon Ahn (CTO) is an expert in AI technology. Prior to joining VICEVERSA, he was the CTO of VITRUV, which develops AI educational services for students.
Kite Hwang (Creative Director) is a professional in the fashion industry. He has extensive knowledge in fashion. He is bridging the gap between fashion and technology.
Finally, Jay Kim (CEO), has extensive experience in the blockchain industry. Previously he was the CFO of Parameta (a.k.a iconloop) and managed development of new business leveraging blockchain technology. He will explore the project’s potential partnerships and licensing opportunities, and set the direction for how we will integrate the technology to future services and platforms.