Questions fréquemment posées

Yes, SEA.AI systems are equipped with thermal cameras, granting them night vision capabilities. Even in adverse weather conditions, their thermal cameras outperform the human eye. Considering the usual tasks undertaken by a crew – whether helming, navigation, or other duties – SEA.AI systems remain steadfastly vigilant, like watchful ‘barrelmen’ atop the mast. Their consistent vigilance ensures real-time alerts in the presence of potential hazards.

Up to 2 users can use the application at the same time on their devices. It is also possible to take over control even from additional devices. E.g Navstation or tablet inside the boat.

Yes, SEA.AI, with its thermal cameras, can detect objects at the same temperature thanks to differences in the emissivity and reflectivity of materials, and the camera’s ability to detect variations in infrared radiation from different surfaces. 

Each material has its own emissivity, which determines the efficiency with which it emits heat, so that objects similar in temperature appear differently.  

Reflectivity also influences the thermal image, as some materials reflect more of the surrounding radiation.  

Finally, infrared variations on surfaces can be detected, enabling objects of equal temperature to be distinguished. For example, a container and water, although at the same temperature, emit infrared radiation differently, making the container visible. 

Using latest camera technology in combination with artificial intelligence, SEA.AI detects and classifies objects to alert crews early and reliably about objects on the surface of the water, day and night.

A Convolutional Neural Networksis (CNN) is a class of artificial neural network that performs segmentation, classification, and detection for a myriad of applications such as maritime safety. Computer vision analyzes images, and then creates numerical representations of what it ‘sees’ using a CNN.

Thermal cameras, also known as infrared cameras, work by detecting and capturing the infrared radiation (heat) emitted by objects. Unlike visible light, which the human eye can perceive, infrared radiation is not visible to the naked eye. Thermal cameras use special sensors and optics to capture this infrared radiation and convert it into a visible image that represents the temperature differences across a scene.

To train our models, we create representative training and test datasets from our large annotated maritime database, which continuously expands with new images capturing a wide range of scenarios.The training set is used to train our AI models using backpropagation. This is a method which, through mathematical optimization, leads the model to correctly detect objects within an image.We apply various data augmentation strategies during the training process to maximize the models’ generalization capabilities. These strategies involve modifying the training images in different ways to help the models learn more robust features.We also perform hyperparameter tuning to find the optimal values for various variables that control the learning process. This involves systematically testing different configurations to ensure the models achieve the best possible performance.Finally, the test set is used to evaluate the models’ performance by comparing annotations to the models’ predictions. This continuous evaluation helps us assess how well the trained models detect objects in unseen data, ensuring they perform well in real-world scenarios.