Looking for a Tool to Refine Object Recognition AI Model on Vertex AI

[CJG64]

I would like to know if there is a tool that would allow me to refine the image object recognition artificial intelligence model developed on vertex AI that I made. I recognize several objects in an image.
I would like to send thousands of images to this model and depending on the model's response, I can validate the response, or modify it by adding or removing the label frames.
I would also like to be able to display the answers according to their scores.
I don't know if Google offers such a tool or if I have to go through a specialized company.

Thank you for your help.

[Aris_O]

Hi @CJG64,

Welcome to Google Cloud Community.

Vertex AI Data classifying Service enables you to improve an image object recognition AI model by manually classifying thousands of images, which you can then use to retrain your model with the new labeled data.

You can set up instructions for labelers to follow and create custom labeling jobs using the Data Labeling Service. By adding or removing label frames, labelers can validate or alter the model's response. You can also see the outcomes of the labeling task as well as the rankings of each label for a specific image.

After you've finished tagging, you can retrain your model to increase accuracy using the labeled data. Using the AutoML Image Training, you can accomplish this.

Here are some documentation you may use as a reference:
https://cloud.google.com/vertex-ai/docs/samples/aiplatform-create-data-labeling-job-images-sample?_g...
https://cloud.google.com/vertex-ai/docs/samples/aiplatform-create-data-labeling-job-specialist-pool-...
https://cloud.google.com/vertex-ai/docs/datasets/data-labeling-job?_ga=2.175503714.-1392753435.16766...
https://cloud.google.com/vertex-ai/docs/datasets/label-using-console?_ga=2.175503714.-1392753435.167...
https://cloud.google.com/ai-platform/data-labeling/pricing?_ga=2.175503714.-1392753435.1676655686

[CJG64]

If i have been clear, and it s good, U made my YEARS. I see that and i come back. Thank you @Aris_O 

[Aris_O]

Hi @CJG64,

I'm glad to help.