Computer Vision Seminar

Transcription

1 Computer Vision Seminar Spring 2017 Instructor: Micha Lindenbaum (Taub 600, Tel: 4331, Student in this seminar should be those interested in high level, learning based, computer vision. They are expected to prepare their lectures carefully, help each other, and ask questions. To complete the seminar, the student should give one lectures (60 pts each), help at least once to another student in the preparation of the lecture (10 pts), be in class and participate in the discussion (10 pts), and prepare a review of one additional paper (20 pts). A list of topics and papers. The papers given here are suggestions for important papers on the topic. They are also good starting points for the paper selection. 1. ( Roee and Yuri) Commonly used image classification CNNs (double) (a) Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems, pages , 2012 (The ImageNet breakthrough, Alex Net). (b) Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for largescale image recognition. arxiv preprint arxiv: , 2014 (VGG). (c) Christian Szegedy, Wei Liu, Yangqing Jia, Pierre Sermanet, Scott Reed, Dragomir Anguelov, Dumitru Erhan, Vincent Vanhoucke, and Andrew Rabinovich. Going deeper with convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1 9, 2015 (Inception, GoogleNet). (d) Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages , 2016 (Residual). 3. ( Nirit) Network Visualization (a) Matthew D Zeiler and Rob Fergus. Visualizing and understanding convolutional networks. In European conference on computer vision, pages Springer, 2014 (Trying to understand what does a CNN represent) (b) Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Object detectors emerge in deep scene cnns. arxiv preprint arxiv: , 2014 ()

2 4. ( Dan) Proposals (a) Jasper RR Uijlings, Koen EA Van De Sande, Theo Gevers, and Arnold WM Smeulders. Selective search for object recognition. International journal of computer vision, 104(2): , 2013 (Proposals) (b) Philipp Krähenbühl and Vladlen Koltun. Geodesic object proposals. In European Conference on Computer Vision, pages Springer, 2014 (Proposals) (c) C Lawrence Zitnick and Piotr Dollár. Edge boxes: Locating object proposals from edges. In European Conference on Computer Vision, pages Springer, 2014 (Proposals - not given) 5. ( David and Yael A.) Detection (double) (a) Ross Girshick, Jeff Donahue, Trevor Darrell, and Jitendra Malik. Rich feature hierarchies for accurate object detection and semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages , 2014 (Detection, RCNN). (b) Ross Girshick. Fast r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, pages , 2015 (Detection, Fast RCNN) (c) Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster r-cnn: Towards real-time object detection with region proposal networks. In Advances in neural information processing systems, pages 91 99, 2015 (Detection, Faster RCNN - not given) (d) Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask r-cnn. arxiv preprint arxiv: , 2017 (Detection, last RCNN version) (e) Joseph Redmon, Santosh Divvala, Ross Girshick, and Ali Farhadi. You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages , 2016 (YOLO). 7. ( Nadav) Optimization for deep Neural Networks 8. ( Oran) Segmentation (a) Jonathan Long, Evan Shelhamer, and Trevor Darrell. Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages , 2015 (Fully convolutional) (b) Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos, Kevin Murphy, and Alan L Yuille. Semantic image segmentation with deep convolutional nets and fully connected crfs. arxiv preprint arxiv: , 2014 (CRF)

3 (c) Bharath Hariharan, Pablo Arbeláez, Ross Girshick, and Jitendra Malik. Hypercolumns for object segmentation and fine-grained localization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages , 2015 (Using multiple layers for segmentation - not given) 9. ( Yehuda) Image processing with deep networks, including Super-resolution and Compression (a) Chao Dong, Chen Change Loy, Kaiming He, and Xiaoou Tang. Image super-resolution using deep convolutional networks. IEEE transactions on pattern analysis and machine intelligence, 38(2): , 2016 (super resolution) (b) Chao Dong, Yubin Deng, Chen Change Loy, and Xiaoou Tang. Compression artifacts reduction by a deep convolutional network. In Proceedings of the IEEE International Conference on Computer Vision, pages , 2015 (compression) 10. ( Almog) 3D reconstruction from 2D (a) David Eigen and Rob Fergus. Predicting depth, surface normals and semantic labels with a common multi-scale convolutional architecture. In Proceedings of the IEEE International Conference on Computer Vision, pages , ( Gilad, Yohai) Generative adversarial Networks (double) (a) Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. Generative adversarial nets. In Advances in neural information processing systems, pages , 2014 (Learning how to generate images by making two networks that compete) 13. ( Yevgeny) Unsupervised training II (a) Carl Doersch, Abhinav Gupta, and Alexei A Efros. Unsupervised visual representation learning by context prediction. In Proceedings of the IEEE International Conference on Computer Vision, pages , ( Yonathan) Caption generation (a) Kelvin Xu, Jimmy Ba, Ryan Kiros, Kyunghyun Cho, Aaron C Courville, Ruslan Salakhutdinov, Richard S Zemel, and Yoshua Bengio. Show, attend and tell: Neural image caption generation with visual attention. In ICML, volume 14, pages 77 81, 2015 (Attention, RNN) 15. (??) 3D point cloud and graph analysis

4 (a) Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. arxiv preprint arxiv: , 2016 (Representing point clouds) 16. ( Yael Y.) Networks on Graphs (a) Joan Bruna, Wojciech Zaremba, Arthur Szlam, and Yann LeCun. Spectral networks and locally connected networks on graphs. arxiv preprint arxiv: , 2013 (b) Mikael Henaff, Joan Bruna, and Yann LeCun. Deep convolutional networks on graphstructured data. arxiv preprint arxiv: , 2015 (c) Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. Convolutional neural networks on graphs with fast localized spectral filtering. In Advances in Neural Information Processing Systems, pages , 2016 (d) Federico Monti, Davide Boscaini, Jonathan Masci, Emanuele Rodolà, Jan Svoboda, and Michael M Bronstein. Geometric deep learning on graphs and manifolds using mixture model cnns. arxiv preprint arxiv: , ( Itamar) Art Generation (a) Leon A Gatys, Alexander S Ecker, and Matthias Bethge. A neural algorithm of artistic style. arxiv preprint arxiv: , 2015 (Generating Artistic effects) Misc - not for lectures 1. Yaniv Taigman, Ming Yang, Marc Aurelio Ranzato, and Lior Wolf. Deepface: Closing the gap to human-level performance in face verification. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages , 2014 (DeepFace). 2. Saining Xie and Zhuowen Tu. Holistically-nested edge detection. In Proceedings of the IEEE International Conference on Computer Vision, pages , 2015 How to give a good lecture: Start with the given paper and search for other papers that look important. Understand the goals and the main ideas. Confirm the choice of the papers with me. Choose 2+ papers and understand them well. Prepare your lecture and slides carefully, making sure that everybody will understand it. Focus on the following issues: What problem does the paper solve? What were the previous methods and why they are not good enough? What are the main principles? (try to identify and isolate main ideas), What is important about the implementation? What are the limitations of the solution? Show typical results. Do not overload the lecture with details but do not omit important ones. Do not write anything that

5 you you cannot explain on the slides. Give the lecture before your partner at least a week before you give it in class. How to write a reviews - The paper (one) you select should be one that was not given in class yet. Refer tot he same questions addressed in a lecture. One page is enough.