Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Super Resolve Event-based Imaging
Investigate the usability of Single Image Super Resolution (SISR) in event cameras.
Keywords: Super-resolution, deep convolutional neural networks, sparse coding
Event cameras are bio-inspired vision sensors that work radically differently from conventional cameras. Instead of capturing intensity images at a fixed rate, event cameras measure changes of intensity asynchronously at the time they occur. This results in a stream of events, which encode the time, location, and polarity (sign) of brightness change.
They have a very high dynamic range (140 dB versus 60 dB), do not suffer from motion blur, and provide measurements with a latency as low as one microsecond. Event cameras provide a viable alternative (or complementary) in conditions that are challenging for conventional cameras. This student work will investigate super resolution techniques to process a stream of events with application to drones and/or autonomous driving scenarios.
Super resolved intensity imaging generates a visually high-resolution (HR) output from its low-resolution (LR) input. However, this inverse problem is ill-posed since multiple HR solutions can map to any LR input. CNN makes a promising approach to SISR proposing super-resolution convolutional neural network (SRCNN). The technique requires to explore the usability in a stream of events and the different network structures to achieve a balance between performance and speed.
Requirements: Background in computer vision and machine learning - Deep learning experience preferable – Excellent programming experience in C++ and Python
Event cameras are bio-inspired vision sensors that work radically differently from conventional cameras. Instead of capturing intensity images at a fixed rate, event cameras measure changes of intensity asynchronously at the time they occur. This results in a stream of events, which encode the time, location, and polarity (sign) of brightness change.
They have a very high dynamic range (140 dB versus 60 dB), do not suffer from motion blur, and provide measurements with a latency as low as one microsecond. Event cameras provide a viable alternative (or complementary) in conditions that are challenging for conventional cameras. This student work will investigate super resolution techniques to process a stream of events with application to drones and/or autonomous driving scenarios.
Super resolved intensity imaging generates a visually high-resolution (HR) output from its low-resolution (LR) input. However, this inverse problem is ill-posed since multiple HR solutions can map to any LR input. CNN makes a promising approach to SISR proposing super-resolution convolutional neural network (SRCNN). The technique requires to explore the usability in a stream of events and the different network structures to achieve a balance between performance and speed.
Requirements: Background in computer vision and machine learning - Deep learning experience preferable – Excellent programming experience in C++ and Python
Generate a high-resolution (HR) image from a low-resolution (LR) stream of events.
Generate a high-resolution (HR) image from a low-resolution (LR) stream of events.