Full Text Available
Note: Clicking the button above will open the full text document at the original institutional repository in a new window.
Improving region based CNN object detector using bayesian optimization
Using Deep Neural Networks for object detection tasks has had groundbreaking results on several object detection benchmarks. Although the trained models have high capacity and strong discrimination power, yet inaccurate localization is a major source of error for those detection systems. In my work,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
AUC Knowledge Fountain
2018
|
| Subjects: | |
| Tags: |
No Tags, Be the first to tag this record!
|
| Summary: | Using Deep Neural Networks for object detection tasks has had groundbreaking results on several object detection benchmarks. Although the trained models have high capacity and strong discrimination power, yet inaccurate localization is a major source of error for those detection systems. In my work, I'm developing a sequential searching algorithm based on Bayesian Optimization to propose better candidate bounding boxes for the objects of interest. The work is focusing on formulating effective region proposal as an optimization problem and using Bayesian Optimization algorithm as a black-box optimizer to sequentially solve this problem. The proposed algorithm demonstrated the state-of-the-art performance on PASCAL VOC 2007 benchmark under the standard localization requirements. |
|---|