Object Detection for Analysis of Piglets Feeding Behaviour in an Incubator
Student thesis: Master Thesis and HD Thesis
- Jonatan Emil Simonsen
4. term, Vision, Graphics and Interactive Systems, Master (Master Programme)
The piglets in the pig industry have multiple difficulties in the beginning of their life. SEGES Innovation is working on an incubator using artificial teats to reduce the number of piglets dying shortly after birth.
This thesis proposes the usage of object detection by using variations of a Faster R-CNN. The architectures are trained and tested using a custom dataset, which have been gathered in collaboration with SEGES Innovation. The models strive towards detecting piglets using the artificial teats to feed in the incubator, and determine the piglet drinking.
The results shows, that the piglet could be detected drinking with an AP of 39.5% using a threshold of 0.7, and distinguishing between the piglets had an AP of 27.4%.
Overall the models did not achieve to detect a piglet feeding using the artificial teats with high precision.
This thesis proposes the usage of object detection by using variations of a Faster R-CNN. The architectures are trained and tested using a custom dataset, which have been gathered in collaboration with SEGES Innovation. The models strive towards detecting piglets using the artificial teats to feed in the incubator, and determine the piglet drinking.
The results shows, that the piglet could be detected drinking with an AP of 39.5% using a threshold of 0.7, and distinguishing between the piglets had an AP of 27.4%.
Overall the models did not achieve to detect a piglet feeding using the artificial teats with high precision.
| Language | English |
|---|---|
| Publication date | 2 Jun 2022 |
| Number of pages | 65 |
| External collaborator | SEGES Innovation P/S Chefforsker Vivi Aarestrup Moustsen vam@seges.dk Information group |