Image models allow you to analyze images to detect objects. Standard models allow you to use general categories like "wood", "furniture" or "chair" and etc. without being specific about the exact model of the chair. To detect specific objects you need custom algorithms and that is one of the most important long-term benefits of using Bitskout. We discuss in an article about custom models and in our blog post about proprietary data.
But let's come back to the original question...
- To create an image model go to AI Models. The list of models will appear:
2. Press the Add button and the screen to set up the model will appear.
3. Specify the model name and description.
4. Next, choose the type of model - Image or Text.
5. In the next step please choose what do you want to do with images:
For our example, let's choose the option of detecting objects.
6. Next, type into the Text field the object labels you want to detect and press Add button. You can add as many labels as you want.
Note: if you don't add any labels, the model will output all labels that it has found.
- If you are not really sure about the label, use the Get Labels feature with a sample image to see what labels are detected.
- If you want to remove the label, use the Recycle bin button to remove it.
7. Adjust the labels according to the test results. Remember, that you can always edit the model later on depending on the results.
8. Choose the model behavior mode: Pass/not pass or Detection mode.
9. Next, choose content evaluation mode.
10. Press Apply and your model will appear in the list of models.