July 2016 - talks - lists.cs.princeton.edu

Ohad Fried's Pre FPO will take place on Thursday, July 14, 2016 (today) at 2pm in CS 402
by Nicki Gotsis 14 Jul '16

14 Jul '16

Ohad Fried's Pre FPO will take place on Thursday, July 14, 2016 (today) at 2pm in CS 402. The members of his committee are as follows: Adam Finkelstein (adviser)' Readers: Eli Shechtman (Adobe), Thomas Funkhouser; Non readers: Szymon Rusinkiewicz, David Dobkin. Everyone is invited to attend his talk. The talk title and abstract follow below: Title: "Photo Manipulation (The Easy Way)" Outline: With the growing number of images taken every second, and the growing size of personal photo collections, the need arises for more sophisticated yet easy to use photo manipulation techniques. This talk will present several projects that aim to “democratize” image manipulation: allowing sophisticated edits that are accessible to novices. I will start by presenting a novel way to arrange object collections, such as personal photo libraries. Collections are often presented visually in a grid because it is a compact representation that lends itself well for search and exploration. Most grid layouts are sorted using very basic criteria, such as date or filename. We present a method to arrange collections of objects respecting an arbitrary distance measure. Pairwise distances are preserved as much as possible, while still producing the specific target arrangement which may be a 2D grid, the surface of a sphere, a hierarchy, or any other shape. We show that our method can be used for infographics, collection exploration and data visualization. We present a fast algorithm that can work on large collections and quantitatively evaluate how well distances are preserved. Next, I will show a method to remove unwanted distracting element from images. We propose a new computer vision task — “distractor prediction”. Distractors are the regions of an image that draw attention away from the main subjects and reduce the overall image quality. Removing distractors — for example, using in-painting — can improve the composition of an image. We created two datasets of images with user annotations to identify the characteristics of distractors. We use these datasets to train an algorithm to predict distractor maps. Finally, we use our predictor to automatically enhance images. Lastly, I will show a method to modify the apparent relative pose and distance between camera and subject given a single portrait photo. Our approach fits a full perspective camera and a parametric 3D head model to the portrait, and then builds a 2D warp in the image plane to approximate the effect of a desired change in 3D. We show that this model is capable of correcting objectionable artifacts such as the large noses sometimes seen in “selfies,” or to deliberately bring a distant camera closer to the subject. This framework can also be used to re-pose the subject, as well as to create stereo pairs from an input portrait. We show convincing results on both an existing dataset as well as a new dataset we captured to validate our method.

1 0

Fisher Yu will present his Pre-FPO on Monday, July 18, 2016 at 1pm in CS 402
by Nicki Gotsis 13 Jul '16

13 Jul '16

Fisher Yu will present his Pre-FPO on Monday, July 18, 2016 at 1pm in CS 402. The members of his committee are: Thomas Funkhouser (adviser), Vladlen Koltun (reader;Intel), Szymon Rusinkiewicz (reader), Adam Finkelstein (non-reader), Kai Li (non-reader). Everyone is invited to attend his talk. The talk title and abstract follow below: Title: Pixel-Level Prediction of Depth and Semantics in Images Outline: Pixel-Level prediction is a common task in computer vision. Human eyes can assign each pixel 3D properties, such as depth and normal, and semantic meaning, such as object category and boundary. The pixel-level information can help algorithm process online content with finer granularity and help autonomous agents navigate in complicated environment. I will present my research in 3D geometry to predict pixel depth in previously impossible cases and convolutional networks that can improve image segmentation results substantially. I will also talk about my efforts to create large scale image dataset which may create new possibilities in computer vision research. I will first talk about pixel-level depth prediction from a set of images with very small motion. The images are taken from videos recording hand motion when he/she tries to hold the camera still. The scenario is challenging due to very small baseline. We analyze the SfM problem from geometry perspective and find an algorithm that can recover camera motion for most of the cases. Further, a multi-view stereo method robust to depth noise is developed to predict the depth at each pixel. We also show that the recovered depth can be useful for some computational photography applications. To predict the semantics of each pixel, or to do semantic image segmentation, we turn to convolutional networks, which is more powerful than the alternatives. A lot of works have studied how to use the network designed for image classification as basis to solve the segmentation problem. I argue that in order to design a better model for image segmentation, we have to change the basis to make it more suitable for segmentation. Toward this goal, I introduce dilated convolution as an additional building block for semantic segmentation and study its properties. We use dilated convolution to build a deeper network that has much wider receptive field than the original image classification network. Our model has got competitive performance on different datasets. Finally, I will introduce my on-going project, LSUN, which aims to make it possible to explore weakly annotated data to help solve the image segmentation problem. In LSUN project, we have collected orders of magnitude more labeled and unlabeled images than any previous datasets. To reach the scale, we design an iterative method to combine the advantages of classification models and human labeling. We also introduce additional statistical tests along the process so that the final labeled set has controlled precision. Up to now, we are able to collect around 70 million labeled images with about 90% labeling accuracy. Additional experiments show that LSUN dataset is poised to give us more insight in understanding images.

1 0

Sema Berkiten will present her FPO, "Capturing, Processing, and Synthesizing Surfaces with Details" on Thursday, 7/21/2016 at 2:00 pm in CS 402.
by Nicki Gotsis 13 Jul '16

13 Jul '16

Sema Berkiten will present her FPO, "Capturing, Processing, and Synthesizing Surfaces with Details" on Thursday, 7/21/2016 at 2:00 pm, CS 402. The members of her committee are: Szymon Rusinkiewicz (adviser), readers: Adam Finkelstein and Hao Li (University of Southern California); and nonreaders: Thomas Funkhouser and David Dobkin. A copy of her thesis is available in Room 310. Everyone is invited to attend her talk. The talk abstract follow below. Geometry acquisition and processing have become increasingly popular in computer graphics and vision, with demand for high-quality models driven by advances in 3D printing, realistic real-time renderings of 3D avatars in video games, digital libraries for historical objects etc. In this thesis, we focus on techniques to produce and process detailed geometry including acquisition of real world objects, processing and fusing the captured data, and synthesizing new surfaces from existing ones. First, we summarize a 2D acquisition technique called photometric stereo to capture high resolution surface details. As a validation step, we solve a misalignment problem for photometric datasets. After the dataset is validated, the surface normals can be computed using one of the photometric stereo algorithms. In order to decide which algorithm to use, we present a synthetic photometric benchmark to evaluate various algorithms for different scenarios. To produce a detailed surface in 3D, we propose an approach to combine a rough 3D geometry with detailed normal maps obtained from different views. We begin with unaligned 2D normal maps and a rough 3D geometry, and automatically align each normal map to the 3D geometry. Next, we map the normals onto the surface, correct and seamlessly blend them together. We then optimize the geometry to produce a high-quality 3D model. Next, we introduce a semi-automated system to convert photometric datasets into geometry-aware non-photorealistic illustrations of surface details that obey the common conventions of epigraphy (black-and-white archaeological drawings of inscriptions). This system is composed of rectification of the surface normals to correct camera perspective, segmentation of the inscriptions from the background, classification of the inscription based on carving technique, and stylization of the inscriptions in various styles. Finally, we present an algorithm for realistically transferring surface details (specifically, displacement maps) from existing high-quality 3D models to simple shapes that may be created with easy-to-learn modeling tools. Our key insight is to use metric learning to find a combination of geometric features that successfully predicts detail-map similarities on the source mesh, and use the learned feature combination to drive the detail transfer

1 0