Online Shopping is the future. eCommerce throughout the world is booming. We shop every single thing in the world except those products where touch and feel are rather important. For example, I never buy clothes online because of the issue with size or colour not being accurate. Same goes with other products too where the colour and actual perception of the product size is rather unclear. I choose to directly shop from the retail stores.

My wife was recently looking for a new watch. And she had her mind set on a watch that has a tint of rose gold. We went through a lot of models in various eCommerce sites like Amazon, Flipkart and other sites that sell watches. Some of the issues that we had with the product details, description and images are :

• Description and Watch colour is tagged as ‘Rose Gold’ while the picture is not super obvious if it is indeed the same colour due to overexposure or poor quality of the photo.
• Description did not say it is ‘Rose Gold’ while images seemed that it was ‘Rose Gold’.

Essentially there are issues with the information not being complete. This happens not just with watches but evident in other product categories too.

How Machine Learning can help ?

Computer Vision is one of the sub-division within Machine Learning. Computer Vision is all about analyzing the images and extracting the details of the image. We could see the application of this in a wide variety of applications. For example, selfie cameras detecting our image to sharpen or smoothen it or identify the age or the bokeh mode where the camera detects our face and blurs the background. Behind all of this, Computer Vision is the science that powers all of this.

Computer Vision can be applied in the eCommerce space too. For instance, all of the images that are mapped to a product can be run via the Computer Vision algorithm to detect objects and extract the metadata of the image. In this case, Computer Vision can analyze the image of the watch and then extract the various attributes of the watch, including the colour. And once the metadata of the image is retrieved, the same can be matched with the product description or tagging and if there are any differences, the same could be reported and analyzed.

This can help to ensure the erroneous data is removed and provide a better customer experience during the online shopping. It is impossible for humans to manually do this activity, but with Machine Learning and Computer Vision models and algorithm, this can be achieved with relative ease.

Computer Vision is also the technology that is driving the detection of objects in self-driving cars through Deep Learning. There are other wide varieties of use-cases where Computer Vision is widely used and as the technology matures, the implications of applying this technology will be more pronounced across industries.