Image recognition is a fast-growing method within data science. With this technique the focus is on a special type of data type: namely photos. With image recognition you learn to recognize an image algorithm. You do this by offering many images with the correct meaning. This makes it increasingly easier for the algorithm to link the correct meaning to a specific image.

In America, the large real estate platform Zillow is already experimenting with the possibilities to determine house prices through this technology. This algorithm is therefore already able to recognize certain properties of a house that may influence the house price.
An emerging niche within image recognition is the automatic analysis of satellite data. For example, there are companies that in this way detect asbestos suspect roofs or can detect (illegal) renovations to houses.

Also during the Green Amsterdam Project  we looked at how we could use Image recognition for making the city more sustainable. During the preliminary investigation it quickly emerged that data on land use is often incomplete, inaccurate or outdated. Then we set up an experimental method with which we analyzed satellite images through image recognition. The aim was to use this technique to gain insight into the current greenery of Amsterdam. In addition, this technique is also suitable to find new places that may be suitable for further greening, such as flat roofs and vacant sites. An additional advantage is that Image recognition can be fully automated and can be applied on a large scale.

Some applications for which image recognition is used:

  • Asset management (monitoring)
  • Smart farming
  • Visual search
  • Asbestos detection
  • Marketing Intelligence
  • Risk Intelligence
  • Property valuations
  • Solar panel detection and optimization
  • City planning
  • Object recognition

Image recognition has many advantages and is widely applicable. The algorithms are becoming smarter and the costs for employability are decreasing. In the coming months we will come up with new examples in which we have used image recognition to gain insights at Matrixian Group.