Perception Technologies.AMPL experience in perception technologies have provided state of the art solutions.
|
Deep Learning for perception applications
Current technologies take advantage on the availability of powerful GPUs which allows the creation of state of the art Neural Networks which allows to detect and identify patterns in images and point clouds.
Our lab has wide experience in developing Deep Learning techniques that has been used in industry and academia for a wide diversity of applications: Autonomous vehicles perception, quality control, drones technologies are some of the examples where our technology has been successfully tested and implemented.
Our 3D detection software stack is composed of different deep neural networks, which is able to process the information gathered by the most popular sensors used in robotics applications: cameras and LiDARs.
Our lab has wide experience in developing Deep Learning techniques that has been used in industry and academia for a wide diversity of applications: Autonomous vehicles perception, quality control, drones technologies are some of the examples where our technology has been successfully tested and implemented.
Our 3D detection software stack is composed of different deep neural networks, which is able to process the information gathered by the most popular sensors used in robotics applications: cameras and LiDARs.
Image detection
Image detection
|
Semantic segmentation
|
Image re-identification
|
3D Detections
Obtaining 3D measurements can be performed in different ways: using a camera and a laser, using two or more cameras and more recently, using 3D cameras also provide color information.
AMPL has experience using 3D technology, it has been performed real time large objects, calibration and synchronization of multiple cameras and the use of RGB-D cameras for modeling environments.
AMPL has experience using 3D technology, it has been performed real time large objects, calibration and synchronization of multiple cameras and the use of RGB-D cameras for modeling environments.
Monocular 3D reconstruction
|
LiDAR based detection
|
Sensor Fusion
|
3D Reconstruction
The 3D and 2D information is treated in combination with other sources or just in form of point cloud. The lab has proven it usability in different application, such as surveillance, automotive or inspection. Some examples of 3D applications developed by AMPL are:
HD Map Reconstruction
|
SLAM
|
LiDAR Fusion
|
Vision Applied to Robots
|
Computer vision has led to a more flexible use of robots, allowing them to perceive and interpret the environment. The works carried out have focused on the identification of artificial and natural markings, which help in the tasks of navigation and location robot.
Cheaper and reduced hardware size allows growing perception that the algorithms can be executed in real time on the robots. |
|
Sensor Setup
Apart from our software portfolio, AMPL know-how includes wide experience in sensor configuration to select the optimal setup of devices for each specific problem. Moreover, we provide state-of-the-art calibration and synchronization technologies for multi-sensor systems.
Setup Design
|
Calibration
|
Synchronization
|
Localization
Another technology that AMPL has sucessfully developed is Fusing our lidar localization solution with GNSS information; where GNSS data is used to improve localization accuracy in places with less map features and landmarks and to prevent the kidnapped robot problem.
Besides, laser information improves accuracy in places where the map has more features and GNSS higher covariance, or error, allowing twopi's data fusion localization to be used in specifically difficult scenarios for GNSS such as urban.
Besides, laser information improves accuracy in places where the map has more features and GNSS higher covariance, or error, allowing twopi's data fusion localization to be used in specifically difficult scenarios for GNSS such as urban.
Classic Localization
|
Our Approach
|
Multi-Object Tracking
|
Tracking allows to provide consistency to our dertections, we are expert in advanced and complex tracking systems that allows to enhance the detection.
We have succesfull implemented online and offline tracking with reverse time tracking capabilities. The system enhances our approaches, not only in precision, but also in range. |
|
Traffic monitoring
|
Recent advances in Artificial Intelligence have provided powerful tools to ensure high levels of safety for Smart Cities. For traffic monitoring, we have developed deep learning algorithms that can detect and track objects from fixed infrastructure, i.e. cameras mounted on traffic lights, by enhancing current models that cannot predict as accurately as they should. This is due to the variation in size and scale of the objects produced when filming in a top-view perspective.
|
|
Quality Control
|
We have developed different applications for quality control in industry, where we apply our knowledge in perception to provide high quality applications in industrial environments.
|
|