Visual AI to convert Lookers into Bookers

Source: Deep Learning on Medium


AI is no longer an option for any enterprise, it is a necessity for the survival

Source

If you were to book a very normal room through an OTA(Online Travel Agency), which out the following hotel rooms would you choose:

Left | Right

No points for guessing, it has to be the one on the right. Images act as a kind of window for the products that we see online. If we find the images pleasing, we move forward to know more about them. The reverse happens if what we see doesn’t please our eyes.

The presence of good quality images is one of the most important factors that drive the online hotel industry and Airbnb’s example further emphasizes this aspect.

How Airbnb averted Bankruptcy by using Image Aesthetics?

Airbnb, an accommodation-sharing site that lets you host strangers for a sum, is a pretty common name today. Since its inception in 2008, it has grown exponentially. However, everything wasn’t so merry in the beginning, in fact, Airbnb was at the brink of bankruptcy in 2009. The company looked into everything right from their business model to their product suitability, but the actual problem turned out relating to Image aesthetics.

Airbnb hosts a site where people see the images of the available accommodations and accordingly book their stay. However, back in 2009, most of the photos posted were taken by people from their mobile phones with bad quality which were terrible and really bad. It was also reported later that some of the houses represented were these photos were actually very good. Airbnb then started on a spree to replace the low-quality images with their high-resolution counterparts and just as expected, the company began to grow and register profits.

Adding aesthetics to the business, just did the trick for Airbnb.


Images: A vital source of Data

The world today is full of data and images form a significant part of this data. Images and image sequence constitute about 80 per cent of all corporate and public unstructured big data. The images are available in multiple forms like photographs, satellite maps, medical images etc. The image database gets only bigger with every passing second. Facebook receives more than 10 million new photos uploaded every hour, and around 1 billion people use Instagram. We’re capturing more images than ever before in the entirety of human history. One of the primary reasons for this surge in image data is due to the rise of smartphones. With state of the art cameras present in our phones, we virtually carry a ‘personal photographer’ in our pockets.


Importance of Images in Business

As the growth of unstructured data increases, systems are trying to interpret images to convert them into meaningful data. Images provide valuable insights to consumers regarding a product. Images also help to generate more engagement which is vital for processes. Today the first point of contact for a business and its clients is through their website and the images on the website form an essential part of this interaction.

According to a study conducted by the Cornell University School of Hotel Administration, Property name and images are the two main things that drive the online hotel choice. The study also revealed that once the people shortlisted some properties, the images of the property were the deciding factor.

results show that property name and Images drive the online hotel choice

A Picture is worth a thousand words

A good photo can make or break a connection almost instantly, with a potential client. Expedia, an American global travel technology company, in its study, concluded that every good image on a website helps sales conversions for Online Travel Agencies.

Hotel and Real Estate sectors are two common businesses which are heavily governed by the quality and kind of images they display on their websites.

So if you are into a hospitality business, you would want to keep these salient points in mind regarding the images displayed on the websites:

1. Good Quality photographs

Properties with a high-quality carousel of hotel images showcasing the impressive and luxurious amenities have a higher conversion rate than the ones with mediocre photos.

The hotel room on the right is definitely going to grab more eyeballs than the one on the left even though the room charges for both of them is nearly the same.

2. Highlighting the major areas

Gorgeous entrances, clean surroundings and greenery have proven to be a positive factor in attracting people to property.

Source

3. Avoiding irrelevant and Misleading images

Hotels should avoid showcasing such images on their websites that people do not want to see. These may include dark, blurry and low-quality photographs. Also, multiple photographs of a single room don’t give much choice to clients. Shots of an empty reception area give the impression that there’s no one there to greet guests.

empty reception areas hint that there’s no one to greet the guests

Some websites display images of property which actually do not exist. This can lure client once but can prove to be detrimental for the company in the future.

4. Aesthetics

Aesthetics, in the world of art and photography, refers to the principles of nature and appreciation of beauty. Aesthetics can increase the visual appeal of the photographs.

source

Photographs with a consistent colour scheme, similar lighting conditions, a bright and radiant backdrop can have a positive effect on the viewers. Similarly, hotels can highlight their seasonal offerings to attract travellers throughout the year. Snowfall pictures in winters to a refreshing swimming pool image during the hot summer months can enable a hotel to showcase its different facets and hence appear appealing all the year round. Aesthetics enhances the attractive quotient of the images.

5. Variety

Variety is the spice of life and humans love it. It makes a perfect sense to have a variety of images to illustrate each facility in the hotel rather than having multiple images of a single facility. By providing a variety of images, one can give travellers a better understanding of the property and an opportunity to visualise it.


DeepPulse is Redefining Image processing

Image Processing is a collection of Machine Learning algorithms built to perform certain operations on an image, so as to either enhance its properties or to derive meaningful insights from it.

DeepPulse is developing state of art AI use cases by combining Computer Vision, NLP and collective Global intelligence to improve sales and process outcomes for large enterprises starting from Hotel Booking industry. We help the businesses obtain the best, relevant and the apt image for their purpose. Some of the open source Deep Learning libraries used by us include PyTorch, Tensorflow, Caffe etc. Our APIs extracts the smallest details from an image thereby providing the user with valuable and useful insights.

How DeepPulse uses Images to help businesses Boost Sales

Customizing the Search Results

Hotels have millions of photographs in their database. So how do we make sure that the best image catches the traveller’s eye? Well, our Classification Algorithm does that neatly by ranking the images on various parameters:

  • Source and destination city
  • Days and season of travel
  • Duration of travel
  • Traveller type
  • Traveller persona
  • Traveller categorisation
  • Approximate budget
  • Most preferred room type and many more.

Highlighting Hotel’s strengths

Our algorithms can easily extract image aesthetics, attributes and quality of an image. This helps to highlight better images on the hotel’s website.

Source

For the above hotel image, our APIs can easily extract the following information.


Showing Amenities based on Source and destination city

We imagine what might attract a family, a couple or a business traveller and then show them the amenities that appeal to their needs and expectations. For example, a family with children will most likely be looking for a hotel with recreational areas where the kids can play and adults can relax. However, this may not be true for a business traveller who will be looking out for better room facilities and will avoid places where there are a lot of children to avoid being disturbed.


DNA-fication

DeepPulse uses six different DNAs to auto-magically predict the behaviour and characteristics of hotel visitors. This enables the OTAs to cater to what the potential guests are looking for. This also ensures right message at the right time to influence guest’s booking decision.


Room Classification

We use AI to classify and tag millions of images in no time

Our algorithm accurately classifies room type for a given image eg bedroom, Living Rooms etc. This is helpful since the image in which the person is most interested in will be surfaced first as it will have a higher rank as compared to other images in the database.


Landmark Detection

Our classifier ranks images on the basis of landmarks, famous monuments and other important structures around them. This is done to ensure that such features are highlighted in the pictures searched by travellers on a hotel’s website.


Presenting the Best and Filtering out the rest.

Our algorithm also ranks the images based on their quality, subject focus and colours. Thus, the dull and bad images are ranked the lowest while, the images with vibrant colours, higher resolution and harmonious colours are ranked higher.


SEO tags Generation

Since most of the hotel bookings are done online, it becomes imperative that we are able to generate Search Engine optimised tags for the properties.


Conclusion

Every feature of a hotel attracts a different set of audience and this is a very important aspect to be kept in mind during Image classification. Nobody likes to spend money on a property that they cannot see. And if what they see is not good, they will never go for it. At DeepPulse we offer tools for e-commerce websites to easily implement image recognition such as described above to Improve the content management of your e-commerce site.


Want to know more about DeepPulse? Get in touch with our super talented team at ankur@deeppulse.ai.