Make AI Smarter Through Continuously Improving Models
Detect Where and What Each Item of Clothing Is
In 2017, we created our first apparel classification model to help retailers streamline their workflow using tools like automatic tagging. The classification model was designed to accurately identify the most salient article of clothing in an image, since that clothing item is likely to be what retailers wanted to sell. Now, we have a new apparel model to share. Read more to find out how what’s changed.
Why Detection Instead of Classification?
Our apparel detection model is designed to locate and identify several articles of clothing in a photo, rather than the most notable aspect of the overall image. So, if you’re using a photo of Lenny Kravitz wearing a massive scarf (remember that?), the classification model will likely focus on the scarf and may miss other clothing items. Detection models are designed to take note of multiple relevant objects in an image. Thus, the many facets of an outfit can be identified independently using an apparel detection model.
In this photo, the classification model focused on trying to identify the scarf, the largest item of clothing. But it is missing other pieces of the outfit.
In this photo, the detection model was able to spot different articles of clothing, regardless of the size of the item in the image.
Our detection model can also be used to locate where an item of clothing is in a photo of multiple people. This comes in handy for retailers who are trying to tag images containing several items of clothing because it streamlines the labeling process, and saves a lot of time. So, for retail labelers looking at group photos in trepidation- fear not. We are here to reduce the time it takes to get the items of clothing labeled and ready to list on your website. Our accurate descriptions can also improve search engine performance, giving potential customers the chance to find your products.
In this photo, our detection was able to identify the type and location of several articles of clothing and accessories.
What’s Coming Next?
We plan to continue training our apparel detection model so that it can recognize a wider range of clothing items, including clothes from different cultures. We are also expanding the taxonomy to give back more fine-grain details of the clothing items. Log in to try our apparel detection model in the portal or check it out on our website today!