Machine Learning and Computer Vision Reading Group 2017

• Jan. 05, 2017 -- He Zhao
  Paper: Ashish Shrivastava, Tomas Pfister, Oncel Tuzel, Josh Susskind, Wenda Wang and Russ Webb, Learning from Simulated and Unsupervised Images through Adversarial Training. arXiv:1612.07828v1, 2016.


• Jan. 12, 2017 -- Xiaowei Zhang
  Paper: Tianfan Xue, Jiajun Wu, Katherine Bouman and Bill Freeman, Visual Dynamics: Probabilistic Future Frame Synthesis via Cross Convolutional Networks. NIPS, 2016.


• Jan. 19, 2017 -- Li Cheng
  Title: Recurrent neural nets (RNNs) and its applications (slides)


• Jan. 26, 2017 -- Shuang Wu
  Paper: Yonghui Wu et al., Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation. ICML, 2016.


• Feb. 02, 2017 -- Yu Zhang
  Paper: David Joseph Tan et al., Fits Like a Glove: Rapid and Reliable Hand Shape Personalization. CVPR, 2016.


• Feb. 09, 2017 -- Chi Xu
  Paper: Joao Carreira, Pulkit Agrawal, Katerina Fragkiadaki, Jitendra Malik, Human Pose Estimation with Iterative Error Feedback. CVPR, 2016.


• Feb. 16, 2017 -- Liang Hui (guest speaker)
  Title: Vision-based Hand Motion Analysis for Human-Computer Interaction
  Abstract: We address the problem of hand pose and gesture analysis in RGB-D images with convolutional neural networks (CNN) and random forests. Both CNN and random forests are prevailing techniques for human hand and body pose analysis in recent years, while their accuracy is still unsatisfactory. In this talk we present two approaches based on CNN and random forest respectively with improved capability of handling ambiguous hand postures. The first is multi-view CNNs, in which a single depth image is projected to multiple planes, each of which is then sent to a CNN for pose regression and their results are fused to output robust prediction. The second is random forests with suppressed leaves, in which the leaf nodes are optimized to reflect their importance in Hough voting. Various real-time demos are also developed based on these methods to assist human computer interaction.


• Mar. 9, 2017 -- Lakshimi Govindarajan
  Paper: Markus Oberweger, Paul Wohlhart, Vincent Lepetit, Training a feedback loop for hand pose estimation. ICCV, 2015.


• Mar. 16, 2017 -- He Zhao
  Paper: Nguyen A, Yosinski J, Bengio Y, et al., Plug & Play Generative Networks: Conditional Iterative Generation of Images in Latent Space. arXiv:1612.00005, 2016.


• Mar. 23, 2017 -- Shuang Wu
  Paper: Oord A, Kalchbrenner N, Kavukcuoglu K, Pixel Recurrent Neural Networks. arXiv:1601.06759, 2016.


• Mar. 30, 2017 -- Li Cheng
  Online/Early Action Detection, Activity Forecasting and All That


• Apr. 06, 2017 -- Yu Zhang
  Paper: C Wan, T Probst, L Van Gool, A Yao, Crossing Nets: Dual Generative Models with a Shared Latent Space for Hand Pose Estimation. arXiv:1702.03431, 2017.


• Apr. 13, 2017 -- Chi Xu
  Paper: Anastasia Tkach, Mark Pauly, Andrea Tagliasacchi, Sphere-Meshes for Real-Time Hand Modeling and Tracking, ACM Transactions on Graphics. Proceedings of SIGGRAPH Asia, 2016.


• Apr. 20, 2017 -- Connie Kou
  Paper: Aviv Tamar, Yi Wu, Garrett Thomas, Sergey Levine, Pieter Abbeel, Value Iteration Networks. NIPS Proceedings, 2016.


• May. 4, 2017 -- Yunchao Wei (Guest Speaker)
  Title: Object Recognition with Image-level Annotation
  Abstract: Among various levels of supervision information (e.g. labels, bounding boxes and pixel-level annotations), the simplest and most efficient one that can be collected for training is the image-level object category annotation. In this talk, he will introduce his recent efforts on multi-label classification, weakly-supervised object detection and weakly-supervised semantic segmentation. His works achieve some state-of-the-art results on these challenging tasks.


• May. 11, 2017 -- He Zhao
  Paper: Li R, Zeng T, Peng H, Ji S, Deep Learning Segmentation of Optical Microscopy Images Improves 3D Neuron Reconstruction. IEEE Transactions on Medical Imaging, 2017.
  


• May 18, 2017 -- Lakshmi Govindarajan
  Paper: Achal Dave, Olga Russakovsky, Deva Ramanan, Predictive-Corrective Networks for Action Detection. arXiv:1704.03615, 2017.


• May 25, 2017 -- Satyam
  Paper: Andre Esteva, Brett Kuprel et. al., Dermatologist-level classification of skin cancer with deep neural networks.. Nature, 2017.


• June 1, 2017 -- Satyam
  Paper: Peter Kontschieder, Madalina Fiterau, Antonio Criminisi, Samuel Rota Bulo, Deep Neural Decision Forests. ICCV, 2015.


• Jun. 29, 2017 -- Shuang Wu
  Paper: Shiyu Huang, Deva Ramanan, Expecting the Unexpected: Training Detectors for Unusual Pedestrians with Adversarial Imposters. arXiv:1703.06283, 2017.


• Jul. 13, 2017 -- Robby Tan (Guest Speaker)
  Title: Rain Removal and Optical Flow in Rainy Scenes
  Abstract: Rain degrades visibility of a scene, causing many outdoor computer vision algorithms to break down. Rain streaks exhibit specular highlights and occlude the scenes. They also accumulate and form atmospheric veiling effects that wash out the hues of the scene and reduce contrast. This talk will discuss two methods dealing with these two types of rain (rain streaks and rain-streak accumulation). One of the methods is based on layer decomposition using Gaussian Mixture Models as constraints, and the other method is based on deep learning. In the deep learning method, new rain models and new architecture are proposed. Aside from enhancing visibility, an optical flow method that works robustly in rainy scenes is also introduced. This method does not assume any rain removal prior process.


• Jul. 27, 2017 -- Li Cheng
  Paper: S. Yuan, Q. Ye, Bjorn Stenger, Siddhant Jain, T-K Kim BigHand2M Benchmark Hand Pose Dataset and State of the Art Analysis. CVPR, 2017.


• Aug. 3, 2017 -- Yu Zhang
  Paper: Christian Zimmermann, Thomas Brox Learning to Estimate 3D Hand Pose from Single RGB Images. arXiv:1705.01389, 2017.


• Aug. 10, 2017 -- Chi Xu
  Paper: Shih-En Wei, Varun Ramakrishna, Takeo Kanade, Yaser Sheikh Convolutional Pose Machines. arXiv:1602.00134, 2016.


• Aug. 17, 2017 -- He Zhao
  Paper: Han Zhang, Tao Xu, Hongsheng Li, Shaoting Zhang, Xiaogang Wang, Xiaolei Huang, Dimitris Metaxas StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks. arXiv:1612.03242.


• Aug. 24, 2017 -- Shuang Wu
  Paper: Joseph Redmon, Ali Farhadi Better, Faster, Stronger. arXiv:1612.08242.


• Aug. 31, 2017 -- Zhenguang Liu
  Paper: German Ros, Laura Sellart, Joanna Materzynska, David Vazquez, Antonio M.Lopez The SYNTHIA Dataset: A Large Collection of Synthetic Images for Semantic Segmentation of Urban Scenes,CVPR16.


• Sep. 07, 2017 -- Huilin Zhu
  Paper: Gao Huang, Zhuang Liu, Laurens van der Maaten, Kilian Q. Weinberger Densely Connected Convolutional Networks, CVPR17.


• Sep. 14, 2017 -- Guosheng Lin
  
  Title: Learning Deep Networks for Semantic Segmentation .
  Abstract: Semantic segmentation is a fundamental task for visual scene understanding. In this talk I will introduce our recent work on learning deep networks for high-resolution semantic segmentation and weakly supervised learning from web images for semantic segmentation.


• Sep. 28, 2017 -- Xavier Bresson (Assoc Professor from NTU)
  
  Title: Convolutional Neural Networks on Graphs.
  Abstract:Convolutional neural networks have greatly improved state-of-the-art performances in computer vision and speech analysis tasks, due to its high ability to extract multiple levels of representations of data. In this talk, we are interested in generalizing convolutional neural networks from low-dimensional regular grids, where image, video and speech are represented, to high-dimensional irregular domains, such as social networks, telecommunication networks, or words' embedding. We present a formulation of convolutional neural networks on graphs in the context of spectral graph theory, which provides the necessary mathematical background and efficient numerical schemes to design fast localized convolutional filters on graphs. Numerical experiments demonstrate the ability of the system to learn local stationary features on graphs.


• Oct. 05, 2017 -- Nastaran Okati
  Paper: German Ros, Laura Sellart, Joanna Materzynska, David Vazquez, Antonio M.Lopez Geometric deep learning on graphs and manifolds using mixture model CNNs, arXiv:1611.08402.


• Oct. 19, 2017 -- Ajay Vishwanath
  Paper: Serena Yeung, Olga Russakovsky, Greg Mori, Li Fei-Fei End-to-end Learning of Action Detection from Frame Glimpses in Videos.


• Nov. 9, 2017 -- Li Cheng
  Paper: Some Recent updates on Generative Adversary Nets (GANs).


• Nov. 16, 2017 -- Ajay Vishwanath
  Master thesis: Real-time Hand Action Detection from Ego-centric Depth Sequences.


• Nov. 30, 2017 -- Zhenguang Liu
  Paper: Fujun Luan, Sylvain Paris, Eli Shechtman, Kavita Bala Deep photo style transfer.,arXiv:1703.07511v3.


• Dec. 07, 2017 -- Shuang Wu
  Paper: Dushyant Mehta et al., VNect: Real-time 3D Human Pose Estimation with a Single RGB Camera. SIGGRAPH, 2017.


• Dec. 14, 2017 -- Nastaran Okati
  Paper: Ashesh Jain, Amir R. Zamir, Silvio Savarese, Ashutosh Saxena, Structural-RNN: Deep Learning on Spatio-Temporal Graphs. CVPR, 2016.


• Dec. 21, 2017 -- Xiaoting Wang
  Paper: Stephan R. Richter, Zeeshan Hayder, Vladlen Koltun, Playing for Benchmarks. ICCV, 2017.


• Dec. 28, 2017 -- Valentina Bellemo
  Paper: Gul Varol, Javier Romero, Xavier Martin, Naureen Mahmood, Michael J. Black, Ivan Laptev, Cordelia Schmid, Learning from Synthetic Humans. CVPR, 2017.



• Reading Group 2016
       
• Reading Group 2015