A reliable autonomous vehicle shall be swift in decision making protecting the passengers and the driver in hazardous situations and to ensure a calm and full of joy ride. Processing of images to localize objects with bounding boxes is called object detection. As the localization is done using bounding boxes the actual boundaries of objects are captured. However, it offers the processing less computationally heavy task to localize an object in a portion of a second. We offer you three types of bounding boxes: Horizontal, Rotated and Oriented. The oriented bounding box can give hints about in which direction an object is heading to.

We are living and driving in a 3D world. To have an in-depth insight, awareness and full cognition of the world around us, we need to understand and locate objects in three dimensions. We offer 3D object detection annotation both in images and videos using 3D bounding boxes known as Cuboids for your AI algorithms.

A consistent 360° perception of the surrounding is essential for an autonomous vehicle to operate safely and reliably all the time. A combination of laser sensors such as LiDAR can provide a vehicle with this perception with point cloud data. This perception capability is boosted even further by RADAR capturing objects far ahead of the vehicle. They do not have the shortcomings of optical cameras for instance in poor weather conditions. To design an AI algorithm to be capable of localizing objecting using laser data, we provide you with accurate, and inexpensive 3D bounding box and 3D point-wise segmentation.

We as humans have several senses to shape our understanding of the environment around us by fusing the output of sense with each other. Vehicles like us need to combine all their perceiving sensors to capture the moment correctly in order to make the right decision. However, the annotation of each sensor data separately, not only multiplies the required effort, but also can lead to inconsistencies between the annotations of the same object in the different sensor outputs. In Torno, we have designed an efficient AI-powered sensor fusion tool that only one sensor output has to be annotated and the annotation of the rest of sensors are created automatically according to positional location. This not only decreases the required effort and costs by several times, but also it brings consistency to the final output.

In an autonomous vehicle, it is required to foresee hazardous situations and to predict the future at least by some golden seconds to prevent the material damage or endangering the pedestrians. This goal can be achieved by detection of each object and tracking it by keeping its object identity. This will allow us to predict the future if an object is occluded by another object for a while and will appear afterwards. This will also allow to predict the trajectory of the object and to calculate whether this trajectory will intersect with the vehicle’s trajectory. We offer you single-object tracking (SOT) and multi-object tracking (MOT) annotation using our Human-guided AI annotation tool to deliver you with high-quality and seamless labeled data.

In Advanced Driver Assistance Systems (ADAS), there is a system called Lane Departure Warning. More broadly, an autonomous vehicle should be able to localize itself between lanes and to have an understanding of lane-marking meanings. Like humans learning the meaning of the lane-markings, an autonomous vehicle should be taught to understand the message conveyed by each lane-marking class. In Torno, using our AI-powered annotation tool, we are able to annotate even very tiny lane-markings with their semantics several times faster than manual annotation and yet with consistency and high-quality.

An autonomous vehicle should know where it can drive and where. This will not only help the vehicle to continue the journey and take over if required, but in case of a hazardous situation it may drive to a safe area rather than driving into an static object which endangers the passengers as well.

Single- and Multi-object tracking annotation is difficult and time-consuming in essence. This becomes even more difficult and poses a serious challenge when tracking tiny objects in airborne and satellite data with complex backgrounds, low-resolution, shadows, and occlusions. Having said that, Torno has developed an AI-powered annotation tool to track very tiny objects such as vehicles in when Ground Sampling Distance (GDS) is even very high. The data can be either orthorectified and registered or be annotated in raw format.

Not all of the images are of RGB modularity. In medical imagery, X-ray imagery is often used to see the bone and skeleton of our bodies to diagnosis the disease and identify the tissue causing pain to the patient. Torno annotation tool is equipped with X-ray annotation tool with high efficiency. In addition to X-ray, our annotation tool supports MRI, and CT-Scan.

In case the localization of relatively thick objects is required, we provide annotation with polygons for pixel-wise semantic segmentation in various medical imaging sensors such as X-ray, CT-scan, and MRI.