Inspirest models on the metaverse by learning in-depth the decision, choice, thinking, and design characteristics of stylists, fashion designers, and artists living from the past to the present. Basically, Fashionic AI under FaaS aims to create AI-based clones of fashion stylists. Thanks to Fashionic AI, users will be provided with a continuous decision support mechanism in the choice of clothes during the shopping process. This will bring a whole new aspect to the Fashion world. Artificial intelligence models will enable the creation of designs that exceed the limits of imagination and the stylists to produce outside the laws of physics and independent of time. Inspirest develops unique deep learning techniques for Artificial Intelligence clones of fashion designers.
One of the most popular usages of Deep Learning is classification. There has been much study on classification in Fashion Analyzes. Classification clothing according to predefined styles using Deep Learning Models is one of the most popular methods. CNN(Convolutional Neural Network) or derived models are most commonly used in fashion style classification with images. Especially models like VGG, Xception, and Inception, which are proven to work successfully on Image Classification, are widely preferred. Style prediction using these models can reach 85% accuracy [1,2].
Another classification method is attribute classification. Items are classified according to predefined attributes, and attributes are each item’s characteristics and objective qualities. Each item can have multiple attributes; that is why it is usually a multi-label task. For example, a dress (category) can have color, pattern, material, price, texture, style, etc., as attributes. CNN and derived CNN models are primarily used in attribute classification with a prediction accuracy of 90% [3].
Classification via Designer Style is a novel area for AI. Fashionic AI will use the best style classification and attribute classification to classify Designer Styles accurately. In Fashionic AI for learning designers, multiple Deep Learning Models like CNN, VGG, Xception, Inception, ResNet will be tested and used to get the best accurate classification model.
Get ready to take fashion to a limitless phase beyond imagination with Inspirest’s Deep Learning techniques.
Referances
[1] Dong, C. Y., Shi, Y. Q., & Tao, R. (2018). Convolutional Neural Networks for Clothing Image Style Recognition. DEStech Transactions on Computer Science and Engineering, cmsms. https://doi.org/10.12783/dtcse/cmsms2018/25262
[2] Takagi, M., Simo-Serra, E., Iizuka, S., & Ishikawa, H. (2017). What Makes a Style: Experimental Analysis of Fashion Prediction. 2017 IEEE International Conference on Computer Vision Workshops (ICCVW), 2247–2253. doi:10.1109/ICCVW.2017.263
[3] Verma, D., Gulati, K., & Shah, R. R. (2020). Addressing the Cold-Start Problem in Outfit Recommendation Using Visual Preference Modelling. 2020 IEEE Sixth International Conference on Multimedia Big Data (BigMM). doi:10.1109/bigmm50055.2020.00043
