Generative AI refers to the AI technology that learns patterns from a vast amount of data and generates new data that has never been seen before. It can generate faces of people who do not exist in the world, write articles in the style of a certain writer, or create new images that describe a certain background or situation. Recently, generative AI has become a hot topic with models such as ChatGPT , Stable Diffusion , Imagen. How can we apply generative AI like ChatGPT in the fashion industry? To find an answer to this question, we asked ChatGPT.
Here’s the first question:
- Generating product descriptions: ChatGPT can be trained on a dataset of fashion product descriptions, and then used to generate new, unique descriptions for new products.
- Personal styling: ChatGPT can be used to provide personalized styling recommendations to customers based on their preferences and body type.
- Customer service: ChatGPT can be integrated into a company's customer service chatbot to provide assistance to customers with their orders, returns, and other inquiries.
- Social media content creation: ChatGPT can be used to generate engaging and relevant content for a fashion brand's social media accounts.
- Trend forecasting: ChatGPT can be used to analyze large amounts of data on fashion trends and predict future trends.
- Virtual try-on: ChatGPT can be used to generate realistic virtual images of clothes on avatars of different body types, so that customers can see how a piece of clothing would look on them before buying it.
- Personalized marketing: ChatGPT can be used to generate highly personalized marketing messages for customers based on their browsing and purchase history.
The essence of ChatGPT is that it is a text generation model based on conversational systems, so using text to generate any content such as creating product descriptions, chatbot services for customer service, and SNS marketing contents, can be a feasible direction. To verify this, we tried a task of generating product descriptions. We used OMNIOUS TAGGER to tag images and inputted the results as the input for ChatGPT, and generated product descriptions based on this.
Integrating ChatGPT with a personalized recommendation system can involve several steps:
- Data collection: The first step is to collect data on user preferences and behavior, such as browsing history, purchase history, and survey responses. This data can be used to train the personalized recommendation system and to understand the user's preferences.
- Preprocessing: The collected data needs to be preprocessed and cleaned to make it ready for the modeling process. This includes tasks such as removing outliers, handling missing values, and feature engineering.
- Modeling: The preprocessed data can be used to train a personalized recommendation system using various machine learning algorithms such as collaborative filtering, decision tree, clustering, and deep learning. The goal is to predict what the user is likely to be interested in based on their past behavior.
- Integration: Once the personalized recommendation system is trained, it can be integrated with ChatGPT. The recommendation system can provide ChatGPT with a list of personalized product recommendations for a specific user. ChatGPT can then generate text such as product descriptions or personalized marketing messages based on the recommended products.
- Evaluation: The entire system should be evaluated using various metrics such as precision, recall, and F1-score to ensure that the generated text is accurate and relevant.
Keep in mind that, ChatGPT is a language model, it generates text given text, but the integration with personalized recommendation system can be achieved by feeding the personalized recommendations to the model as a context.
It's also worth noting that the integration with a personalized recommendation system can help ChatGPT to generate more accurate and relevant text by providing it with more information about the user's preferences and behavior.
Interestingly, even though we did not input style information about the image above, it inferred style information such as "making it perfect for both casual and formal occasions". This shows that by utilizing ChatGPT, it is possible to generate more natural and rich information using data or AI models we have.
However, it is not clear how it can be applied to virtual try-on, where we should generate fake images of certain type of target clothing. Therefore, we asked more questions.
ChatGPT is primarily a language generation model, so its direct application to generating images for virtual try-on may be limited. However, it can still be used in conjunction with other technologies to create a virtual try-on experience.
One way to use ChatGPT for virtual try-on is to generate text-based descriptions of how an item of clothing or accessory might look on a person. For example, ChatGPT could be trained on images and descriptions of clothing and accessories, and then be used to generate text such as "this dress would look great on someone with an hourglass figure," or "these earrings would add a pop of color to any outfit." This text-based description can be used to give customers an idea of how an item might look on them, without having to generate an image.
Another way to use ChatGPT in virtual try-on is to use it as a part of a chatbot that can assist customers in finding the right clothing and accessories for them, by answering their questions and making suggestions based on their preferences and body type.
Additionally, GPT-3 can be integrated with computer vision models to generate images. This can be done by fine-tuning GPT-3 on the text-image pairs and with the help of GPT-3, a computer vision model could be able to generate images of the similar style or pattern on a given text input.
It's worth noting that, although ChatGPT can be integrated with other technologies for virtual try-on, it would likely be just one part of a larger system that would also involve computer vision and other technologies to generate images.
Here, we can observe that ChatGPT's power is not limited to simply generating plausible answers to given sentences or questions, but it can easily be integrated with various other AI technologies. With help of ChatGPT, we can easily switch context through language, a common interpretable communication tool, and through this, we can utilize achievements in other AI fields that have been actively researched so far. One example may be in the field of personalized recommendations, or it may be a virtual-try on technology that combines computer vision and graphics technology. The possibilities are endless. Depending on how we deliver our context to ChatGPT with the appropriate prompt, we can use ChatGPT as a tool for expanding into various fields.
Of course, there are also skeptical views on the widespread use of generative AI like ChatGPT. Actually referencing non-existent papers as if they were published in some conference or generating seemingly plausible answers without any proof of what is true can be an ethical issue. Recently, Yann LeCun shared concerns about the fundamental limitations of ChatGPT, which is trained with RLHF (Reinforcement Learning with Human Feedbacks), saying that while it's a great writing assistant, it may not be suitable for expansion as a search engine. Also, International Conference on Machine Learning (ICML) recently announced that they are going to prohibit any usage of any LLM-generated text in their academic papers. One thing is certain, the possibilities are open depending on how we use it.
In this article, we looked at how the recent hot topic of generative AI, ChatGPT, can be used in the fashion industry from the perspective of ChatGPT. Many applications that incorporate ChatGPT are already being released (reference: vscode, chatGPT-for-google, youtube video summary), and Microsoft has also presented the concept of LMOps to better utilize LLM (Large Language Model). Microsoft Azure has also announced the introduction of services that utilize ChatGPT. In this flow, the near future will be a year where the benefits that AI can bring to the fashion industry as a whole can grow further. And we hope that OMNIOUS.AI can also contribute to this flow.