Hello everyone!
Zia intelligently validates the image fields of a record based on training images you've provided, which helps reduce the amount of time users spend manually validating images. Zoho has now made some enhancements to this feature, with the major one being the addition of the match and detect option.
Match and detect images for comprehensive validation
Let's say we're using Zia to validate product images on an ecommerce platform.
We use the match option to ensure that these images match the color, design, and overall quality of the reference images. For example, a seller uploads an image of a red dress and Zia validates it based on how closely it matches the reference images used for training.
The detect option helps sellers validate product images where the product forms a part of the image. Say you want to upload product images of sofas, and you want the images to contain a sofa in a room. When a seller uploads such an image, Zia will try to detect the sofa within that image.
Now, Zoho has added the option to do both in a single rule. This improves the accuracy of validation, as an image will be approved only if both of the following are present:
- A complete match to reference images
- Detected (or undetected) objects
For example, let's say you want to match the product images of laptops but also ensure that the webcam, screen, keyboard, trackpad, and so on are present as well. By using this method, you can ensure that tablets, desktops, and other objects are not mistakenly approved.
Decide how Zia should get suggestions for learning
Previously, reviewers could approve images which Zia had declared invalid and help Zia by identifying its mistakes. Reviewers can now manually or automatically suggest those images to train Zia. This is available for rules that use custom images for model training.
Snapshot view of training images
The snapshot view gives you a quick visual breakdown of the training images related to a rule.
Remove images from your training data
When you click View Images in the snapshot view, you'll have the option of removing training images individually or in bulk. To make this easier, you can now filter images by label/object and time range.
Approve suggestions to training images
You can also add any suggestions your reviewers submit for your training data; just select the label for the approved image and mark the objects detected in it.
To learn more about this feature, please refer to
Zia Vision.
![](https://static.zohocdn.com/zoho-desk-editor/static/images/file.png)
Availability: Enterprise edition and above
DC: US DC