Neurotechnology, a provider of deep learning-based solutions and high-precision biometric identification technologies, today announced the release of a new version of its SentiSight.ai image labeling and recognition platform. The new release includes some additional features and an improved interface for free platform users. It also includes a number of powerful features, such as object detection model training, offline models, project sharing and labeling time-tracking, which will be available for paid users.
SentiSight.ai is a web-based platform that can be used for image labeling and for developing AI-based image recognition applications. It has two major goals: the first is to make the image annotation task as convenient and efficient as possible, even for large projects with many people working on image labeling, and the second is to provide a smooth and user-friendly interface for training and deploying deep neural network models. The ability to perform both of these tasks on the same platform provides the advantage of being able to label images and then train and improve models in an iterative way.
“SentiSight.ai has become a platform of choice for image labeling and almost any AI-related task,” said Dr. Karolis Uziela, SentiSight.ai team lead from Neurotechnology. “It has also become one of the first such platforms to offer the ability to download offline models, which allows our clients to be completely independent both from the platform and from their connection to the internet.”
The SentiSight.ai platform was first released as a free-to-use tool on November 19, 2018. Since then, a number of new features have been added to the free version:
- Improved image annotation tool. The original version of SentiSight.ai only allowed adding classification labels, bounding boxes and polygons. Now users can also label points, polylines and bitmaps. Bitmap labeling speed can also be significantly increased by using the smart labeling tool, which allows users to select a few points in the foreground and the background and let the AI extract the labeled object. The labeled images can be directly used for model training on the SentiSight.ai platform, or they can be downloaded and used for in-house model training.
- Pre-trained models. Originally, SentiSight.ai focused on providing a user interface for custom model training. Now it also provides several pre-trained models that can be used out-of-the-box without any additional training. These pre-trained models can be used for several tasks, such as content moderation, goods classification, automatic hashtags, people counting and more.
- Similarity search. This new ready-to-use SentiSight.ai feature allows users to upload an image and find all similar images to this query in their data set. It also allows users to perform NvN similarity searches in their data set where all similar image pairs are retrieved.
In addition to the above features that are available for all customers, the new version of SentiSight.ai has several features that will be available for paid customers:
- Object detection model training. Previously SentiSight.ai offered only classification model training. These types of models can be used to identify what is inside the image as a whole. Now SentiSight.ai also offers object detection model training. This type of model can not only identify objects in an image but also predict their exact location.
- Offline models (free 30-day trial available). In the previous version of SentiSight.ai, the image recognition models could be used either via REST API or web interface. Both of these options required internet connection. The new SentiSight.ai offers a third option: to download and use the image recognition model offline. An offline model can be downloaded as a free 30-day trial after which the user has an option to buy a license. The price of the license depends on the speed of the model, and it is a single time payment.
- Shared labeling projects and time tracking. To make large annotation project handling easier, SentiSight.ai allows a project to be shared among multiple users so that multiple people can label images in the same project. The project manager can quickly filter and review the images labeled by a particular project member, track each person’s progress and time spent on labeling, as well as manage user roles and permissions.