Image Classification

In Intelligent Document Processing, you can easily identify the type of the document by just looking at the document instead of referring the textual context. For example, in case of KYC documents like PAN, Aadhaar, or License, you can easily recognize the document. For such set of documents, the objects in the document image or the structure of the image varies. Such documents can be easily categorized by using image classification. 

Image Classification helps to train models on images. These image datasets are very large, and you have to train the model on the Smart Bot, so it is recommended to have these datasets on the Smart Bot in a pre-configured location. And during Model training, you need to provide the dataset/folder  name from the location. 

Image classification training takes more time as compared to Text Classification, however, the run-time prediction with image classification is faster as it works on images and there is no OCR involved to extract tests from the images.

To open an Image Classification application, navigate to Smart Bot > Classification > Image Classification.

The Image Classification page displays two tabs namely Dataset and Model.

• Dataset: Dataset page helps to add dataset location/ configured dataset folder name from the Smart Bot. 

The Datasets section is displayed with the following details:

• Dataset Name: Specifies the name of the dataset.
• Description: Specifies the description of the dataset.
• Folder Name: Specifies a name to identify the dataset.
• Created Time: Specifies the date and time of the dataset when it was created.
• Created By: Specifies the name of the user who has created the dataset.
• Actions: Use the Actions column for performing the following operations: 
  • Download:  To download the uploaded dataset.
  • Delete:  To delete the uploaded dataset.

       

• Model: Model Page helps to select required dataset to train a model.

Model page displays following information:

• Classifier Name: Specifies the name of the text classifier model.
• Dataset Name: Specifies the name of the selected dataset.
• Status: Specifies status of the classifier model.
  • Completed:  The model is trained successfully.
  • In Progress: The training is in progress.
  • Failed: the model failed to train.
• Description: Specifies the description of the classifier.
• Last Modified Time: Specifies the latest date and time when the text classification model was trained.
• Accuracy: Specifies the accuracy value of the trained model. The value is between 0 and 1. Higher value indicates higher confidence in the prediction. Accuracy helps choose the correct model.
• Actions: Use the Actions column for performing the following operations: 
  • Delete:  To delete the uploaded dataset.
  • Download:   To download the uploaded dataset.
  • Predict:  To predict images.