function readOnly(count){ }
Sign Up ›
Starting November 20, the site will be set to read-only. On December 4, 2023,
forum discussions will move to the Trailblazer Community.
Learn More

Browse by Topic
ShowAll Questionssorted byDate Posted
+ Start a Discussion
Sukanya BanekarSukanya Banekar 

Is it possible to update trained dataset in einstein API

Is it possible to update trained dataset in einstein API?
dsiebolddsiebold
If you created a dataset from a .zip file, you can add to that dataset by uploading additional data in a .zip file (see https://metamind.readme.io/docs/create-examples-from-zip). You can then train the dataset again and it will create a new model.
Sukanya BanekarSukanya Banekar
The procedure you have suggested is to create dataset and trained dataset. But if dataset is already trained it is not possible to add images later, I have got the solution. To update dataset continuously don't create dataset from zip follow below steps:
1. Create empty dataset, add labels to it.
2. Keep on adding images to it.
3. Train the dataset.
4. Again add images and keep training it.


 
Sukanya BanekarSukanya Banekar
Also, if we get probality of the image, is it possible to get the image name to which it has got match?
Michael Machado 22Michael Machado 22

Hi Sukanya- You can add more data to an existing dataset before or after you train your model.  After adding new data, you will need to kick off a new training session and that will generate a new model.  This is best done through our zip upload functionality for adding new data to an existing dataset.  https://metamind.readme.io/docs/create-examples-from-zip)

In the future we are adding in feedback loops and other ways to incorporate new data into existing models to retrain an existing model on new data and maintain the same model ID.  
Sukanya BanekarSukanya Banekar
Hi Michael,
Thank you so much for you reply on this.
In document it is mentioned,
User-added image

So the solution was given in the document itself create empty dataset and keep adding images into that and train accordingly.

Thanks, 
Sukanya