Search
he projects aim to develop new algorithms for video objects segmentation in the presence of noise. Video object segmentation is one of the most studied problem in the computer vision community. The approaches have been shown to be very successful in video segmentation of rigid objects like cars, semi-rigid objects like people or surgical tools. However the contributions related to video segmentation of objects with irregular shape changing in time are very limited. The example are not limited but often concerns bio-medical objects like (1) objects whose appearance change in time, both color and shape like teeth structures during the dental treatment, i.e. cavity ideally should disappear during the process; (2) the objects whose observed shape change in time due to the large motion/displacement in 3D e.g. the fast motion of sperm tail observed in microscope Objective: In this project we attempt to combine the general knowledge about the objects and their dynamic with knowledge learned from spatial-temporal data to segment objects with irregular shapes from video data corrupted by noise/artifacts in the case of multiple uncertain ground truth labels. Results: open-sourced implementations of algorithms, datasets, publications in top computer vision journals and conferences (140-200 pkt on ministry list).
Details
- Project's acronym:
- LearnNoisyVideo
- Financial Program Name:
- OPUS
- Organization:
- Narodowe Centrum Nauki (NCN) (National Science Centre)
- Agreement:
- UMO-2024/53/B/ST6/04273 z dnia 2025-01-29
- Realisation period:
- unknown - unknown
- Project manager:
- dr inż. Anna Jezierska
- Realised in:
- Department of Biomedical Engineering
- Request type:
- National Research Programmes
- Domestic:
- Domestic project
- Verified by:
- Gdańsk University of Technology
seen 1 times