University of Twente Student Theses
Instance Level Semantic Part Segmentation in Cows
Abdul Cader Hasanain, Mohamed Asif Hassan (2019) Instance Level Semantic Part Segmentation in Cows.
|
PDF
24MB |
| Abstract: | Segmenting a cow’s body into its constituent parts namely semantic part segmentation task provides a rich part segmentation that can be used for applications such as activity analysis, cow identification and pose estimation, etc. The goal of this thesis work is to achieve instance level semantic part segmentation in cows. To attain this goal, we propose to use Mask R-CNN as the baseline algorithm for cow parts detection and segmentation. Given a part-level annotation, Mask R-CNN outputs initial part predictions and segments. However, this task is challenging as some parts of the cow such as the limbs are similar in shape and size, which makes it difficult to differentiate between them. On performing experiments with Mask R-CNN, we observe that the algorithm fails to differentiate between the limbs of cow. To tackle this issue, we propose Post-processing methods and End-to-End training models to obtain accurate part predictions. In order to facilitate this research, four different datasets that can detect and segment cows, cow parts, key-points and keymasks are created. These datasets consist of the same images but are annotated for different objects and tasks. Evaluation is done on all the proposed methods and experimental results demonstrate the effectiveness of these methods. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Embedded Systems MSc (60331) |
| Link to this item: | https://purl.utwente.nl/essays/79304 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page