How could i use Ai Center to make a custom Ml Model That extract adress from my given data (adhar cards , a unique citizen identification card in india) , no matter where is position is of the fields ,
How could i do that ? and does the labeling in datalableing actauly only cover that specific area just how the form extractor do in UiPath studio or does the lableing in ai center just get that coordinates as a refrence but its not static or fixed position it just search for the data around and insides that specific area of selection ?
is there any type of training where the ai model itself just try to identify data points and categorized the data on itself without even us to give them a schema of data points or coordinates
if second one is possible it will really helpful as i will provide too many data that i have , yes i dont have any issue in correcting the data if the ai model is getting wrong data , but i want that it should clusterr or say categorized the data itself even getting the correct lables of the data itself , i dont want to provide any information to the model
You can make this work using AIcenter & Document Understanding. First you need to create the a AI center project , then create a data labelling session where you can label ID documents for the training , please ensure you train a good amount of samples and then in ML package section select generic document understanding model and finally run a pipeline using the labelling dataset and with the selected package.
Now you can create your ML skill which you can use in your DU workflow to extract relevant information from the ID documents.
You can follow the same steps as mentioned here in this post:
One thing to take care of is while selecting the Ml Package you have to select Out of the box Packages> UiPath Document Understanding> DocumentUnderstanding. select latest version, submit, give details.
Now in Data Labelling session you would have to create the fields from scratch, choose appropriate data type while creating fields.
And the rest of the steps are same, (you won’t have to import a schema, or enter a prelabelling URL this time), you would have to mark the fields manually for the first time.
If you retrain it later, you can use the Public URL for ML skill you created along with DU API key and it would be able to predict the data.
What if i dont want to provide any labeling at all , i just want to give a feedback of is the data is correct or not in some rating systems like 1to 5 where 1 is worst 5 mean most accuarte or exact match
and the ml model itslef categorize the data gives the label after process many data