We have discontinued our cloud-based data annotation platform since Oct 31st. Contact us for private deployment options.
"Autonomous shipping is the future of the maritime industry. As disruptive as the smartphone, the smart ship will revolutionize the landscape of ship design and operations", said Mikael Makinen, president of Rolls-Royce's marine division.
Over 1,000 ship accidents occur annually in shipping industries, leading to billions in economic losses and endangering crew safety. However, up to 96% of maritime accidents involve human error, according to Allianz statistics. As the maritime industry continues to embrace innovation for greater efficiency and sustainability, autonomous shipping powered by AI promises a safer future. But first, high-quality training data is key to developing intelligent models like neural networks that will steer unmanned vessels. This post explores how maritime data annotation is charting the course for reliable, efficient autonomous marine transportation.
The global autonomous ships market size is projected to surge from USD 5.21 billion in 2022 to USD 9.87 billion by 2030, exhibiting a CAGR of 8.4% during this period as reported by Fortune Business Insights. Driving this growth are players like Wärtsilä, Rolls-Royce, and Mitsui O.S.K. Lines (MOL) develop automated navigation solutions. With enhanced connectivity, sensors, and onboard decision-making capabilities, the future of shipping is undoubtedly AI-enabled.
Autonomous Systems for Smart Shipping in the Maritime Industry
Despite the progress in technology, the maritime industry still grapples with human errors, which account for 75% to 96% of shipping insurance losses or USD 1.6 billion. Factors like poor visibility, adverse weather patterns, fatigue due to long working hours, and insufficient rest contribute to these errors.
The growing digitalization of the maritime industry and the improved connectivity of vessels are shifting more decision-making capabilities onshore. Therefore, minimizing human errors through automation and AI algorithms becomes crucial for increased efficiency and safety in the industry's continued growth.
Automation serves as a key enabler for autonomous solutions, replacing human effort with machines. However, for autonomous systems to become a reality in the maritime sector, three broad capability areas and enablers are critical:
Situational Awareness: Gathering wide range data about the vessel's surroundings using sensors like radars, lasers, and cameras.
Decision Making and Logic: Intelligent AI models interpreting data to decide a safe and effective course of action in real time.
Action and Control: Control systems that enable autonomous actions, executing decisions accurately and safely, typically handled by humans, such as maneuvering the vessel, adjusting speed, etc.
Levels of Autonomous Shipping for Maritime Vessels
Similar to autonomous vehicles, according to the definition of the International Maritime Organization (IMO), an "autonomous surface vessel at sea" is a vessel that can operate, to varying degrees, independently of human interaction. IMO classifies ship automation as:
Vessel with automated processes and decision support for onboard crew
Remote controlled vessel with humans onboard
Remote controlled vessel with no humans onboard
Fully autonomous vessel that can make and execute all needed decisions.
Various vendors have proposed a range of sensing solutions for assistive, autonomous, and remote operations. These sensors accurately measure precise local and global positions and provide situational awareness and hazard detection for automatic control of the vessel.
In addition, all sensors provide outputs that enable seamless integration into the navigation system of the ship, whether it is automated manipulation, remote control, or a fully autonomous navigation system. Context awareness is built on the AI techniques of lidar, millimeter wave radar, and camera devices.
Advantages of Intelligent Marine Shipping
The future of the maritime shipping industry is becoming increasingly intelligent, which brings unprecedented opportunities in terms of improving efficiency, ensuring safe operation and achieving zero emissions.
Increased Safety Through Autonomous Shipping: Human-caused ship collisions will not only lead to economic losses caused by shipping delays, but also cause serious consequences such as casualties and dangerous goods leakage. With the continuous optimization of designed autonomous driving systems, the risk of accidents will be significantly reduced, providing higher safety for ocean shipping.
Energy Efficiency Improvements With Smart Ships: Intelligent ships have intelligent navigation and energy efficiency management functions, and use artificial intelligence and big data analysis technology to process and analyze the information obtained by the perception system to optimize the design of routes and speed. Real-time monitoring of ship status, loading status and fuel consumption can timely adjust the navigation strategy, significantly improve the energy efficiency management level of the ship, thereby reducing energy waste.
Reduced Maintenance Costs Through Autonomous Shipping: At present, ship maintenance mainly relies on highly specialized technicians, which is costly and has a long maintenance cycle. However, with the improvement of intelligence level, technologies such as autonomous decision-making and fault self-diagnosis of ships will gradually solve problems such as the shortage of maintenance talents, effectively reduce maintenance costs, and improve maintenance efficiency.
How Maritime Data Annotation Enables AI Models for Smart Ships
To unlock these benefits, autonomous vessels need AI and computer vision models trained on diverse, high-quality data relevant to marine environments. This is where maritime data labeling becomes critical.
Sensor Fusion is Key for Smart Ships
As autonomous vessels rely on fusing data from multiple sensors, you need an annotation platform capable of handling complex multi-sensor datasets. BasicAI Cloud*, our AI-powered data annotation platform, offers advanced capabilities specifically for maritime sensor fusion scenarios. Our intuitive interface allows syncing and unified tagging of camera, LiDAR and 4D radar (beta) streams. Annotators can switch seamlessly between 2D, 3D and temporal views to label objects and environments accurately. Customizable workflows help adapt to diverse data types and formats.
Specifically for LiDAR point clouds common in autonomous ships, BasicAI Cloud* provides powerful 3D annotation tools. You can conduct semantic segmentation in point clouds to classify areas like shore, piers, open water etc. Object bounding boxes can be drawn precisely in 3D space based on point cloud geometry. Granular point-wise tagging is also available to label individual points. Maritime sensor fusion demands specialized data annotation capabilities. With BasicAI Cloud*'s purpose-built multi-sensor support, you can create high-quality training datasets to develop perception algorithms for safe autonomous navigation.
Ship Image and Video Annotation for Target Detection
With BasicAI Cloud*, you can effectively annotate various sea objects in images and videos. These can range from ferries, buoys, boats, speedboats, kayaks, sailboats, swimmers, seaplanes, and more. The platform also allows for the annotation of state attributes such as motion/stationary along with environmental factors like weather, water flow, and location.
The annotated data then helps train AI models to identify crucial elements in the scenario, providing necessary information for the autonomous control of the ship.
Data Annotation for Navigable Area Detection Model Training
BasicAI Cloud* further empowers you to annotate navigable areas for autonomous navigation. Similar to lane markings for self-driving cars, you can draw poly-lines to define the channel boundaries for ships. Our platform's 3D polygon annotation tool can be used to mark areas such as the shore and piers where boats are not allowed to navigate.
These annotations form the basis of training data for smart ships, helping AI models identify safe navigable areas, and further enhancing the safety of autonomous shipping.
Full Maritime Data Annotation Services for AI Models
For large-scale maritime data annotation projects, developing accurate AI models to enable autonomous ships requires massive volumes of labeled training data. At BasicAI, our professional data annotation teams have the skills and capacity to deliver high-quality maritime dataset creation at scale. With over 7 years of experience annotating millions of data assets, we provide end-to-end services tailored to your project needs:
Scalable annotation workforce trained for the maritime domain
Secure proprietary annotation platform with customizable workflows
Active project management and QA to ensure labeling accuracy
Consulting to determine optimal annotation strategies and project plans
Our maritime data experts understand real-world navigation and can annotate complex multi-sensor data like radar, sonar and thermal feeds. We leverage proven workflows honed from past autonomous vehicle projects involving camera, LiDAR and radar data fusion.
By partnering with BasicAI for your maritime data annotation needs, you gain access to domain knowledge, robust quality practices and rapid turnaround at scale. Our annotated datasets become the fuel to train performant AI models for reliable autonomous marine navigation.
Contact BasicAI Experts to Create Custom Datasets for Your Maritime AI Annotation Projects Today!
* To further enhance data security, we discontinue the Cloud version of our data annotation platform since 31st October 2024. Please contact us for a customized private deployment plan that meets your data annotation goals while prioritizing data security.
*Note: The sensor data examples used in this article are sourced from the Pohang Canal dataset (CC BY-NC 4.0), which provides multimodal marine data collected in the restricted waters of Pohang City. The sensor suite consists of three laser scanners (one 64-channel lidar and two 32-channel lidars), a marine radar, two visual cameras used as a stereo pair, an infrared camera, a 6-direction all-around view camera, an AHRS and a GPS RTK. The dataset includes sensor calibration parameters and baseline trajectories based on SLAM. It was collected along a 7.5 km route comprising narrow canal areas, inner and outer harbor areas, and offshore areas. The goal of the dataset is to facilitate research into autonomous piloting for vessels.