Pianese, Alessandro (2020) Benchmarking SOTA visual object detection techniques for medical applications. Master's Internship Report, Computing Science.
|
Text
mCS_2020_PianeseA.pdf Download (1MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (116kB) |
Abstract
Recently several deep-learning based object detection models (i.e. Faster R-CNN, SSD, YOLO,...) have obtained very good results in well known large-scale object detection benchmarks (i.e. PASCAL VOC, COCO). We propose to benchmark these methods for object detection with real-life medical datasets, such as a collection CT scan slices. The above deep networks have proven to work and obtain good results with huge dataset but it is unknown if they can still perform well with a real-life dataset which is considerably smaller. Furthermore, real-life datasets are also unbalanced with rare diseases appearing less than common ones resulting in lower classification accuracy. This resulted in models needing to be heavily modified to perform on such data sets.
| Item Type: | Thesis (Master's Internship Report) |
|---|---|
| Supervisor name: | Petkov, N. |
| Degree programme: | Computing Science |
| Thesis type: | Master's Internship Report |
| Language: | English |
| Date Deposited: | 26 Nov 2020 12:20 |
| Last Modified: | 26 Nov 2020 12:20 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/23631 |
Actions (login required)
 |
View Item |