Enhancing image segmentation with point prompt augmentation
This project aims to make use of point prompt augmentation techniques to enhance and improve the image segmentation performance of the Segment Anything Model 2 (SAM2). Using SAMAug as a framework for point augmentation strategies, different point generation methods were used when training on the Vec...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/181188 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | This project aims to make use of point prompt augmentation techniques to enhance and improve the image segmentation performance of the Segment Anything Model 2 (SAM2). Using SAMAug as a framework for point augmentation strategies, different point generation methods were used when training on the Vector-LabPics dataset. The report will explore these techniques (random sampling, entropy-based sampling, distance based sampling) as well as evaluate the performance of the baseline SAM2, against SAM2 with data augmentation incorporated during training. |
|---|