The interactive environment of SentiSight.ai is designed for training deep-learning models and provides these capabilities:

• Image labeling toolkit – allows attaching labels to images for image classification, object detection and image segmentation models. An intuitive interface makes labeling faster and easier. Output labels are automatically saved in a format suitable for deep-learning algorithms.
• Training environment – a model can be trained on the prepared images without any coding via the intuitive user interface.
• Pre-trained models – some image recognition models are ready to be used out-of-the-box without any training. These models include general classification, places recognition and nudity detection.
• Interactive statistics – information about the model's performance is produced after the training process. Prediction accuracy, precision, recall and many other metrics allow users to measure their models' performance. This information can be immediately viewed and filtered to briefly state how efficient the model is as well as give guidance on its improvement.
• Online model use – all models trained by a user can be employed to make predictions based on new, previously unseen images. The models can be used online inside the SentiSight.ai platform or via a REST API.
• Similarity Search – users have a possibility to use a similarity search tool that compares images. A user can perform either 1-to-many similarity search, which retrieves all images similar to the query image, or many-to-many similarity search, where all similar image pairs are found in the data set. This tool is ready to use out-of-the-box, without any training.

A custom project can be ordered if a task seems to be "non-standard" or rather complicated. In this case our experts will take care of the model's setup and training. The user just needs to take care of image labeling.

See the SentiSight.ai website for more information.