ADVIT (Analysis of Data-like Video Image and Text) is the most advanced deep learning platform used to create and manage high-quality training data all in one place. ADVIT includes deep learning models for automated labeling and ease of annotation.

The following are the premier features of ADVIT:

1. Custom UI - Custom user interface (UI) designed specifically for the ease of manual annotation and automated labeling.
2. Secure - Easy to label data, securely export data in multiple standard formats, swiftly build and train models, and integrate these models into real-time workflows.
3. Models Included – ADVIT includes 10 Deep learning (DL) models which can automatically label data across more than 100 classes.
4. Integration - Capable of integrating custom-trained models for annotation.
5. Tool to annotate at scale: With efficient Data management for projects running simultaneously, multiple teams across departments can work concurrently to increase overall productivity.
6. Enterprise Identity & Access Management: ADVIT provides identity and access management hence making it easily scalable across teams and locations. Clients can custom integrate this product as per need and make the entire process collaborative for prompt communication within the team.
7. Data Support - Supports all type of data
   1. Image (.jpg, .jpeg, .png)
   2. Video (.mp4, .avi, .mov)

Secure Dataset management platform for labeled and unlabeled data

ADVIT is protected with multiple layers of security measures to prevent unauthorized ccess to your projects, data, models, and annotations.

1. Registration:
   The first time you register on ADVIT, you are required to enter your username, email id, and password to register.

1. Login:
   
   1. Enter your Admin Username and Password.
   2. For additional security, Admin can determine which parts of the project each user has access to, and also limit the number of login attempts. By default, after six failed attempts, the account is locked and the user must wait for a few minutes before trying again.
   3. Locked accounts can be reset by the Admin.
2. Sign out of the ADVIT:
   
   1. In the upper-right corner, click the Account (Account) icon.
   2. Click on Log Out.

The Sign In page displays a message that you are logged out. We recommend you sign out of the Admin any time you leave your computer unattended.

By clicking on the project you want to work on, you will be directed to the dataset dashboard. The dataset page displays all the datasets uploaded for the respective project. The client can add multiple datasets (video format) by clicking on the dataset button present in the top right corner of the screen.

To navigate to the Manual Annotation page, click on Manual Annotation under the “Annotate” tab.
The Instructions for Annotating an image is given as follows:

1. Enter the file ID in the ‘Enter file ID’ box and click on the ‘Go’ button.
2. Select the desired annotation shape from the ‘Select Shape’ Dropdown list.
3. Choose the category from the ‘Category’ Dropdown list and If you feel like adding new categories to the dataset go to the ‘Category Page’ from ADVIT Menu and follow the given steps.
4. Click on the ‘Add’ button and start annotating the object from the image.
5. After Annotating the object, click on the ‘Save’ button to Save the annotated object

1. The Annotation will be featured in the image as Shown Above.
2. The ‘Delete’ button is utilized to Delete the Annotation.
3. The ‘Update’ button will Update the Annotation on basis of ‘Category’ and ‘Shape’.
4. The Co-Ordinates Column records the annotation points of the Coordinates and lists accordingly. You can delete the co-ordinate points and redo annotation.
5. Finally, click on the ‘Save all objects to server’ Button which will save all the annotated objects from the image to the server and will go to the next image for annotation.

Use AI to help you build AI.

ADVIT makes it easy to deploy and repurpose existing models.

Automated Data-labeling uses your model predictions to label your training data. This reduces the time and efforts for the data-labeling and increases the efficiency of the data-labellers.
This tutorial demonstrates how automated data-labeling using Machine Learning (ML) models can be configured onto your system using the ADVIT UI / API. With automated data-labeling, you can use model predictions to label your inputs. Automated data-labeling can help you to prepare training data, or assign other useful labels and metadata to your inputs. Since models are doing most of the work of annotating your data, this enables you to speed up and scale up your annotation process while ensuring quality standards, typically reducing the human effort of labelling data by orders of magnitude. This process is built into our APIs and it seamlessly integrates with all the functionality of the ADVIT platform.

The following ML/ DL models are available in the ADVIT model library and are ready to integrate with your existing models.

1. Man, Woman detection
2. Adult detection
3. Child detection
4. Face detection
5. Facial sentiment analysis

ADVIT is a well-advanced Deep Learning platform with different types of annotations like Bounding box, Polygon, cuboidal, etc. Along with the Automated Bbox annotation, ADVIT now has the capability of Automated Polygon Annotation. Data Scientists can now integrate custom Deep Neural Network models in ADVIT to annotate data at scale. Data Scientists can also download the data in the desired format (post annotations) like Coco(.json) etc. Automaton AI also has the capability of integrating pre-trained Segmentation Models using ADVIT inferencing API.

The instructions to begin with Auto Polygon Annotation are as follows:

Once all the above processes are completed, we can begin with the actual annotation.

Above shown image is the Layout of the Attribute Assignment Page.

1. Enter the desired file ID in the ‘Enter File ID’ box and click on the ‘Go’ Button.
2. Below, you can get the Statistics of Attributes assigned as well as Attributes yet to be assigned with ‘Attribute Key’.
   
3. For Assignment of Attribute, Select the Annotation from the image.
4. ‘Attribute Key column shows you a list of Attributes. Select the Attribute from the list.
5. ‘Attribute Value’ column shows you a list of Attribute Values. Select the Value among them or Select the ‘CD’ value if you cannot define an attribute to the Annotation.
6. Click on the ‘Save’ Button and the Attribute value will be assigned to the annotation shown in the below image.

Automaton AI’s ADVIT now allows you to create and train models on the platform itself. You can add the model’s name, model version, select appropriate model architecture, select the Deep Learning framework of your choice, set hyperparameters & parameters, and finally, select the dataset and categories you want to train your model on. In addition to annotating your data, you can also configure and train the DNN models in ADVIT with just a few clicks, thus creating an all-in-one platform.

The steps to train a model are as follows:

Once the data scientists add a model to the serving, they can download the models’ dockers as well. With the click of a single button, the downloaded doctored model will be ready to be deployed. The downloaded model then can directly be used for inferencing or as the user needs.

To download Inference docker:

ADVIT now supports the Mini Exception model for model training. This model can be used for use cases such as facial sentiment analysis. This pre-trained model supports seven categories which are as follows: Happy, Sad, Neutral, Disgust, Angry, Surprise, Fear.

To make use of the pre-trained model:

Data Scientists can now integrate their own custom Deep Neural Network models using the ADVIT APIs. This functionality helps users to check the accuracy of the model and also can auto-label the datasets which are added by the user into ADVIT. Users can also export the annotations from the ADVIT in standard formats like Coco(.json) etc. The Data Quality Analysis team can perform the quality checking of the annotations done through the model and can also add missing annotations to the database if any.

The steps to integrate your custom DNN model are:

Prerequisites:

This feature in ADVIT allows data scientists to monitor the real-time status of the model training. A few of the parameters which are monitored are learning rate, regression loss, classification loss, learning rate Decay, mAP (if applicable), etc. Based on the statistics, a data scientist can take a run time call whether the training is to be continued or stopped and make necessary changes in parameters and hyperparameters in the next model training version.

Users can re-train the model and maintain multiple versions to achieve optimal accuracy. It will help data scientists to keep track of different versions of the specific model and can also differentiate the trained data. Various model versions can also be maintained to check the accuracy of a particular dataset. One can set different frameworks, parameters & hyperparameter values to differentiate the model accuracy over several datasets.

Once the model is trained, data scientists can evaluate the model by using the weights stored for multiple epochs or with the latest/best weights. The results for evaluation are then downloaded automatically for ease of visualization. This can help the data scientist for a quick glimpse of how the model is performing on unseen datasets.

The model can be evaluated using the following steps:

ADVIT has new functionality that will help data scientists compare the model versions. The page will show a tabularized version of characteristics like model type, framework, training accuracy, training Loss, model accuracy, no. of epochs, etc. This will help data scientists to select the best-performing model version among all versions of the specific model.

To compare the different model versions:

After the model training and evaluation are done, the next step is Adding a model to serving. This feature will logically segregate best-performing models from the rest of the pool. It will help data scientists to use these models for further uses such as Auto annotation, Evaluation, and Inferencing.

Once the model is trained, users can compare the versions and choose which best serves their purpose. Once finalized, users can click on the 'Add model to Serving' option present for the respective model.

When your project is first deployed, a default Admin account is created with login credentials that give you full administrative access. As a best practice, you should create another user account with full Administrator access. That way, you can use one account for your everyday administrative activities and reserve the other as a “Super Admin” account. This can be helpful if you forget your regular credentials, or they somehowbecome unusable.

If there are others on your team or service providers who need access, you can create a separate user account for each and assign restricted access based on their role. To limit the modules that users can access when they login to ADVIT, you must first create a role with limited scope and only the necessary resources selected. Then, you can assign the role to a specific user account. Admin users who are assigned to a restricted role can seeand change data only for associated modules, but cannot make changes to any global settings or data.

For users or roles that are temporary, you can also set an expiration date for the user account. Default User types for ADVIT:

1. Super Admin
2. Admin
3. Automaton AI user
4. Annotator
5. Project Lead

Create a new user

Assign a User Role

The shortcuts ease the process of annotation as well has helps in quality check of the annotations.