Purpose of the project:
- Build an intelligent platform to create and label image data. The system supports labeling by learning from labeled data (AI-in-the-loop), expanding the labeled data set with the support of intelligent learning algorithms. The system also allows synthesizing and generating additional data on new domains that are lacking or do not have data.
- Develop new algorithms for important and popular problems in Computer Vision.
- Build large and semantically labeled data sets for problems in Vietnam in 3 areas: (i) traffic video analysis; (ii) analysis and diagnosis from medical images; (iii) generate descriptive sentences from photos in the Vietnamese context.
- Develop new solutions for practical problems in Vietnam in 3 areas: (i) traffic data analysis; (ii) analysis and diagnosis on medical images; (iii) query big data from daily life logs and activities.
Results:
International competitions:
- 5th place in the AI City Challenge 2020 (track Traffic Anomaly Detection), held during the CVPR 2020 conference
- 10th place in AI City Challenge 2020 (track Vehicle Counts by Class at Multiple Intersections), held during the CVPR 2020 conference
- 1st place in SHREC 2020 (track Retrieval and Classification of Relief Surface Patches), held during the EuroGraphics 2020 conference
Data set:
- Built a data set about anomaly events in daily life with events such as floods, fires, falling trees, traffic jams, traffic accidents… Currently the data set is available over 10,000 photos and continues to be added and expanded.
- Built a dataset to detect tumor types in intestinal endoscopy images based on the BioMedia 2019 Development dataset and KVASIR dataset v2 (4470 images with 4944 subjects).
Implementation:
A platform serves to label and generate related visual data
- Building the architecture and basic features for the platform to serve labeling and generating related visual data
Some secondary results (in addition to the list of registered products):
- 02 articles accepted to be published at the ACM International Conference on Multimedia Retrieval ICMR 2020 (Rank A2 according to Qualis)
- 01 article accepted for publication at CVPR Workshop 2020
- Guide students to achieve 1 First Prize, 1 Second Prize, 1 Third Prize for national scientific research students (Eureka 2019) in the field of Information Technology
