Shipping detector demo
Image recognition is a relatively new development in the maritime sector. It is increasingly applied for collecting information, automatic sailing or to prevent accidents from happening. The detection of maritime objects is possible with a variation of sensors such as color or 360 camera images, radar, LiDAR, thermal, multi-beam and sonar data. For example we can detect the status of vessels within harbors, or automatically map certain routes and directions. For image recognition, it doesn’t matter if its yachts, sloops, fisher boats or even recreational vessels, not a single object will sail past our detection. We can make specific measurements under and above water in lakes, canals, rivers and the sea. To show some of the possibilities of image recognition on a normal image with the use of artificial intelligence we have made an online demo where you can test a certain image to see how the standard algorithm works for different vessels within the sea, inland and fishing maritime sectors.
Common Objects in COntext
The image that is shown above is trained on the COCO dataset and it calculates the opportunity that an object is a vessel within the image. The learning algorithm that we use here detects different kinds of vessels such as: cruise ships, container boats, sail boats, yachts, Royal Navy boats, sloops and even kayaks. The system can be trained further than the demo to detect the specific vessels, situations or other maritime objects.
Common Objects in COntext
The image that is shown above is trained on the COCO dataset and it calculates the opportunity that an object is a vessel within the image. The learning algorithm that we use here detects different kinds of vessels such as: cruise ships, container boats, sail boats, yachts, Royal Navy boats, sloops and even kayaks. The system can be trained further than the demo to detect the specific vessels, situations or other maritime objects.
Image recognition with deep learning
For this example we have used an algorithm that is trained on a thousand photos with different vessels that come from the COCO dataset (Common Objects in Context). This database includes images of boats where you can exactly see where the maritime object is located in the photo. Therefore, a learning algorithm can train itself to automatically recognize different watercraft in different pictures. Off course you can also use this technique to detect other objects as well. For example we can also detect different fish, that can be helpful for fishermen to see where to place the nets and bait. We use tools like Keras and TensorFlow to build our deep learning networks and to see how well the algorithms perform.
