Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Retrieval Robust to Object Motion Blur
Fast moving objects are defined as objects that move over significant distances over exposure time of a single image or video frame. Thus, they look significantly blurred. Detection, tracking, and deblurring of such objects have been studied in recent years. However, there are still no methods for robust retrieval of such objects in large image collections.
Keywords: retrieval, fast moving objects, detection, deep learning
A lot of methods that deal with fast moving objects have been proposed [1], including detection [2], tracking [4], and deblurring [3]. However, a related retrieval task has not been studied before. A general object retrieval [5] is a hot topic in computer vision, but it focuses on general objects, usually landmarks. A method that is robust to object motion blur is still missing.
[1] https://github.com/rozumden/fmo-deblurring-benchmark
[2] FMODetect: Robust Detection of Fast Moving Objects, ICCV 2021
[3] DeFMO: Deblurring and Shape Recovery of Fast Moving Objects, CVPR 2021
[4] Tracking fast moving objects by segmentation network, ICPR 2020
[5] Fine-tuning CNN image retrieval with no human annotation, PAMI 2018
A lot of methods that deal with fast moving objects have been proposed [1], including detection [2], tracking [4], and deblurring [3]. However, a related retrieval task has not been studied before. A general object retrieval [5] is a hot topic in computer vision, but it focuses on general objects, usually landmarks. A method that is robust to object motion blur is still missing.
[1] https://github.com/rozumden/fmo-deblurring-benchmark [2] FMODetect: Robust Detection of Fast Moving Objects, ICCV 2021 [3] DeFMO: Deblurring and Shape Recovery of Fast Moving Objects, CVPR 2021 [4] Tracking fast moving objects by segmentation network, ICPR 2020 [5] Fine-tuning CNN image retrieval with no human annotation, PAMI 2018
The main goal of this thesis is to re-train a state-of-the-art object retrieval method to be robust w.r.t. blur.
The idea is to learn a representation that is robust to blur, and can match a blurred object to the unblurred version of it, e.g. if object query has a blurred object.
The main goal of this thesis is to re-train a state-of-the-art object retrieval method to be robust w.r.t. blur. The idea is to learn a representation that is robust to blur, and can match a blurred object to the unblurred version of it, e.g. if object query has a blurred object.
