Towards dependable object detection
A high confidence in object detection is very crucial for object detector modules to be used in real world applications. Though the confidence scores in object detection can be improved by using better and larger training data set and using more robust architectures, which is an approach from the co...
Saved in:
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/153447 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | A high confidence in object detection is very crucial for object detector modules to be used in real world applications. Though the confidence scores in object detection can be improved by using better and larger training data set and using more robust architectures, which is an approach from the computer vision side, this paper aims to improve the detection confidence by finding the effective viewpoints which makes a dependable use of the available object detectors. In particular, we investigate the effect of viewing distance on detection confidence of an object detector, which can further be used to control a camera mounted on a mobile robot to approach a better viewing position. We consider the cases with both fixed focal length and variable focal length camera. Experimental results are presented to demonstrate the efficacy and validity of the techniques presented in this work. |
|---|