Efficient Object Detection using Fully Convolutional Neural Networks
Thesis by: Bharti Munjal
Supervision: Iro Laina,
Federico Tombari
Abstract
In recent years, Fully Convolutional Networks (FCNs) have set the state of the art in various dense prediction problems, either classification or regression. As part of this thesis, we investigate Fully Convolutional Residual Networks for the problem of object detection in RGB and/or RGB-D images. Our approach combines two-fold information, in order to detect the class and bounding box for objects in the image, given a specified set of known classes. Specifically, we aim to jointly perform semantic segmentation, assigning each pixel in the image to its corresponding class, and offset regression, predicting for each pixel a vector that points to the center of its object, with a unified network. This method additionally holds the potential of being real-time. We will compare the accuracy and efficiency of this approach with other recent object detection methods.
Literature
- Girshick, R., Donahue, J., Darrell, T., & Malik, J. (2014). Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 580-587).
- Girshick, R. (2015). Fast r-cnn. In Proceedings of the IEEE International Conference on Computer Vision (pp. 1440-1448).
- Ren, S., He, K., Girshick, R., & Sun, J. (2015). Faster R-CNN: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems (pp. 91-99).
- Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2015). You only look once: Unified, real-time object detection. arXiv preprint arXiv:1506.02640.
- Dai, J., Li, Y., He, K., & Sun, J. (2016). R-FCN: Object Detection via Region-based Fully Convolutional Networks. arXiv preprint arXiv:1605.06409.