About Custom Models

Custom models on the DeepLobe platform are developed by users or organizations for specific tasks or applications. These models are built using a variety of computer vision and text analytics approaches, and are trained on custom datasets that are tailored to the needs of the task at hand. They are often used in situations where off-the-shelf models are not sufficient for the task at hand, or where the user has specific requirements that existing models cannot meet.

DeepLobe provides a graphical user interface that allows users to build and train models by connecting different blocks or modules together which makes it easy for users who are new to the field or who don't have a lot/any experience with coding.

Data preparation

Transforming raw data from datasets into a format that can be used to uncover insights or make predictions is known as data preparation.

A dataset is a collection of information that can be used to train and test a machine learning model. This data can be anything from a set of numbers to text or even images, and a dataset generally aims to cater to one or more applications. A dataset may be labeled, and therefore, ideal for training and testing supervised models. However, there are also unlabeled datasets that are used to train unsupervised models.

Data annotation

Data annotation tools are the platforms that allow us to tag and label objects from images, videos, and text for the AI models to recognize them and use them to generate predictions.

Let’s get an overview of DeepLobe’s data annotation tool. Once the dataset is uploaded, the images in the dataset need to be labeled (depending on your use case and the custom model that you want to build) for the machines to detect and recognize what is present in the images.

How to improve your custom model’s accuracy?

Feed the model with more data: One of the most effective ways to improve the accuracy of your machine learning model is to train it on more data. Try to collect more data that is representative of the problem you are trying to solve and add it to your training dataset. DeepLobe splits the data into 80:10:10 where 80% of the dataset is used to train the model and 20% is set aside as validation data(10%) and test data(10%) which is used to make predictions against the trained model. You can then compare these predictions against the actual labels given for the inputs and repeat this process for every test dataset.

Data cleaning and preprocessing: Make sure your data is clean, complete, accurate, and consider performing preprocessing steps to improve the quality of your data, such as removing outliers, handling missing values, and normalizing the data.

Outliers are observations in a dataset that are significantly different from the other observations, such as measurement errors, data entry errors, or due to the presence of rare events.

Deeplobe - Custom model subscription