# Visual similarity search

<figure><img src="https://1172273446-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-Lw4gk5ARKsgVHFCdPZw%2Fuploads%2F6mmh4DdR9XYaPdpHz1Aw%2Fvisualsim.png?alt=media&#x26;token=ceb4cd83-857e-4575-9121-f446d228257f" alt=""><figcaption><p>Visual similar properties</p></figcaption></figure>

## How to use

* Input "visual\_similarity": 1 to activate the function.
* Add a custom weight from 0.1 to 1 to determine the impact of sorting the results with 0.5 as default: "weight\_visualsimilarity": 0.5.
* Input an image for the target object via example “url”: “<https://house.image/facade-house.jpg”>.
* Optionally adjust the other weights to improve the sorting of visual similar properties. When only interested in the visual similarity of the reference objects, set the rest of the weights to zero: "weight\_innersurfacearea" : 0, "weight\_buildyear" : 0, "weight\_transactiondate" : 0, "weight\_distance" : 0.

{% hint style="info" %}

* Visual similarity can only be calculated for objects of which a facade image is known in the database.
* The performance of the API when activating “visual\_similarity” is slower because of the activation of the model to parse the images and find visual similar properties.
  {% endhint %}

## Results

* The reference objects are sorted based on the calculated weight
* Per reference object, additional information is returned:
  * Image: The facade image of the reference objects is added per reference object when known: “Image”: “<https://house.image/house.jpg”>.
  * Visual similarity score: “VisualSimilarityScore”: 82.425.