Now we all like online shopping, cheap and convenient, is lazy people’s choice. However, online shopping has the risk ah baby! Because you can’t clearly judge whether a dress is suitable for you as a physical store try-in, online shopping can only rely on model photos for “brain-replenishing”, which is a great test of your figure and eyes. Sometimes wear on the model body “fairy” clothes, to their own body, carelessly became a “seller show”, there is wood?
Little friends, have you ever had a similar painful lesson when you bought clothes online?
Like this kind of slug.
Or this kind of slug.
But don’t worry, recently, Adobe online an AI virtual fitting black technology, you can preview the virtual mannequin on any clothing, never afraid to become a “seller show”
Over the past decade, it’s obvious that platforms that allow shoppers to virtually try on cosmetics, clothing and accessories have become popular, venturebeat reported. According to a survey by klarna, a banking firm, 29 per cent of shoppers prefer to browse items online before they actually buy, while 49 per cent are interested in measuring a solution so they can determine whether a product is appropriate before they buy.
Based on this idea, a team of researchers from Adobe, the Indian Institute of Technology and Stanford University explored what they called “image-based virtual try-through” technology called SieveNet. It maps a garment to a virtual object, preserving the features of a garment, including wrinkles and folds, without blurring or seeping the texture.
SieveNet transfers costume images to virtual models
The purpose of SieveNet is to take images of clothes and mannequins and, while preserving the original body shape, posture and other details, to generate new images of the models wearing the clothes. To achieve this, it uses multi-stage techniques, including deforming clothing to align it with the posture and shape of the mannequin, and then transferring the deformed texture to the model.
The authors of a paper detailing the work point out that geometric warping takes into account changes in shape or posture between clothing images, as well as masking in model images (for example, long hair or cross arms). The dedicated module in SieveNet can predict the level of rough conversion and fine-grain correction on top of previous rough conversions, while another module can calculate rendered images and masks on a mannequin.
In experiments using four Nvidia 1080Ti graphics cards on 16GB of RAM, the researchers trained SieveNet in a data set that included about 19,000 images of frontal female models and top product sms. They reported that in qualitative testing, the system handled masking, posture changes, seepage, geometric warping, and overall quality retention better than the baseline. In addition, it achieves state-of-the-art results on qualitative indicators, including the Fr?chet Start Distance (FID), which takes photos from the target distribution and the evaluated system (in this case SieveNet), and uses AI objects to identify systems that capture important features and preserve similarity.
SieveNet is not the first to eat crabs, to be exact.
In 2019, French beauty giant L’Oreal said it would offer lipstick virtual testing technology to US e-commerce giant Amazon through its augmented reality (AR) and artificial intelligence company ModiFace. “With this new ausit-backed virtual experience, Amazon customers can easily try thousands of lipsticks and share them with friends in mobile photo albums without having to worry about buying the wrong color number,” said Nicolas Le Bourgeois, head of Amazon’s beauty division. “
The system of startup Vue.AI has learned to create realistic postures, skin tones and other features by analyzing the characteristics of clothing. From a snapshot of the garment, it can generate model images of various sizes, 5 times faster than traditional photo shoots!
Gucci and Nike also offer apps that allow people to try on shoes.
Gucci launches AR app to allow customers to virtually “try on” Ace sneaker series
The researchers assert that systems like SeiveNet can be more easily integrated into existing applications and websites. “Virtual try-on – visualizing fashion products in a personalized environment – is particularly important for online fashion transactions because it compensates for the lack of direct in-store shopping experiences,” they wrote, “and we have significant improvements over the current state-of-the-art method of image-based virtual wear.” “