Tasks

[yhalk]

Preparation

Everyone who can start preparing now, should install Keras, TF, select a dataset and get a network running on it. (Thanks to Sen for papers and datasets)

Datasets

http://benchmark.ini.rub.de/?section=gtsdb&subsection=dataset

https://hci.iwr.uni-heidelberg.de/node/6132

http://cvrr.ucsd.edu/vivachallenge/index.php/traffic-light/traffic-light-detection/

http://www.cvlibs.net/datasets/kitti/

https://www.cityscapes-dataset.com/

http://robotcar-dataset.robots.ox.ac.uk

https://github.com/udacity/self-driving-car

https://www.kaggle.com/sudipdas/pedestriandataset

• Whoever chooses one, put a note so that we all work on different datasets.

Papers

ImageNet Classification

• Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems. 2012.

• He, Kaiming, et al. "Deep residual learning for image recognition." CVPR 2015.

Segmentation / Object Detection

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (2015), S. Ren et al.

• Fully convolutional networks for semantic segmentation (2015), J. Long et al.

• Redmon, Joseph, et al. "You only look once: Unified, real-time object detection." arXiv preprint arXiv:1506.02640 (2015).

• L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, and A. L. Yuille. "Semantic image segmentation with deep convolutional nets and fully connected crfs." In ICLR, 2015.

Visual Tracking

• Wang, Naiyan, et al. "Transferring rich feature hierarchies for robust visual tracking." arXiv preprint arXiv:1501.04587 (2015).

• Wang, Lijun, et al. "Visual tracking with fully convolutional networks." Proceedings of the IEEE International Conference on Computer Vision. 2015.

End-to-end Driving

• https://devblogs.nvidia.com/parallelforall/deep-learning-self-driving-cars/

• https://arxiv.org/pdf/1604.07316v1.pdf

[yhalk]

Team task for w/c Aug 14

• Train Faster-RCNN (forked in Boris' branch) with the German traffic dataset (http://benchmark.ini.rub.de/?section=gtsdb&subsection=dataset)

• Prepare a ROS module for taking pictures from Kinect, saving them and make them ready for feeding to the model

• Deploy on the Jetson/Turtlebot

• Print some images of traffic signs and see the model in action

[mocialov]

Lets do the maze task. Example:

[mocialov]

Successfully applied frcnn to 1) http://benchmark.ini.rub.de/?section=gtsdb&subsection=dataset Next step: 1) https://github.com/udacity/self-driving-car 2) http://cvrr.ucsd.edu/vivachallenge/index.php/traffic-light/traffic-light-detection/