We propose Deep Feature Interpolation (DFI), a new data-driven baseline for automatic high-resolution image transformation. As the name suggests, it relies only on simple linear interpolation of deep convolutional features from pre-trained convnets. We show that despite its simplicity, DFI can perform high-level semantic transformations like "make older/younger", "make bespectacled", "add smile", among others, surprisingly well—sometimes even matching or outperforming the state-of-the-art. This is particularly unexpected as DFI requires no specialized network architecture or even any deep network to be trained for these tasks. DFI therefore can be used as a new baseline to evaluate more complex algorithms and provides a practical answer to the question of which image transformation tasks are still challenging in the rise of deep learning.

Scene understanding algorithms in computer vision are improving dramatically by training deep convolutional neural networks on millions of accurately annotated images. Collecting large-scale datasets for this kind of training is challenging, and the learning algorithms are only as good as the data they train on. Training annotations are often obtained by taking the majority label from independent crowdsourced workers using platforms such as Amazon Mechanical Turk. However, the accuracy of the resulting annotations can vary, with the hardest-to-annotate samples having prohibitively low accuracy.

Our insight is that in cases where independent worker annotations are poor more accurate results can be obtained by having workers collaborate. This paper introduces consensus agreement games, a novel method for assigning annotations to images by the agreement of multiple consensuses of small cliques of workers. We demonstrate that this approach reduces error by 37.8% on two different datasets at a cost of $0.10 or $0.17 per annotation. The higher cost is justified because our method does not need to be run on the entire dataset. Ultimately, our method enables us to more accurately annotate images and build more challenging training datasets for learning algorithms.

Recognizing materials in real-world images is a challenging task. Real-world materials have rich surface texture, geometry, lighting conditions, and clutter, which combine to make the problem particularly difficult. In this paper, we introduce a new, large-scale, open dataset of materials in the wild, the Materials in Context Database (MINC), and combine this dataset with deep learning to achieve material recognition and segmentation of images in the wild.

MINC is an order of magnitude larger than previous material databases, while being more diverse and well-sampled across its 23 categories. Using MINC, we train convolutional neural networks (CNNs) for two tasks: classifying materials from patches, and simultaneous material recognition and segmentation in full images. For patch-based classification on MINC we found that the best performing CNN architectures can achieve 85.2% mean class accuracy. We convert these trained CNN classifiers into an efficient fully convolutional framework combined with a fully connected conditional random field (CRF) to predict the material at every pixel in an image, achieving 73.1% mean class accuracy. Our experiments demonstrate that having a large, well-sampled dataset such as MINC is crucial for real-world material recognition and segmentation.

Natural illumination from the sun and sky plays a significant role in the appearance of outdoor scenes. We propose the use of sophisticated outdoor illumination models, developed in the computer graphics community, for estimating appearance and timestamps from a large set of uncalibrated images of an outdoor scene. We first present an analysis of the relationship between these illumination models and the geolocation, time, surface orientation, and local visibility at a scene point. We then use this relationship to devise a data-driven method for estimating per-point albedo and local visibility information from a set of Internet photos taken under varying, unknown illuminations. Our approach significantly extends prior work on appearance estimation to work with sun-sky models, and enables new applications, such as computing timestamps for individual photos using shading information.

The appearance of surfaces in real-world scenes is determined by the materials, textures, and context in which the surfaces appear. However, the datasets we have for visualizing and modeling rich surface appearance in context, in applications such as home remodeling, are quite limited. To help address this need, we present OpenSurfaces, a rich, labeled database consisting of thousands of examples of surfaces segmented from consumer photographs of interiors, and annotated with material parameters (reflectance, material names), texture information (surface normals, rectified textures), and contextual information (scene category, and object names).

Retrieving usable surface information from uncalibrated Internet photo collections is challenging. We use human annotations and present a new methodology for segmenting and annotating materials in Internet photo collections suitable for crowdsourcing (e.g., through Amazon’s Mechanical Turk). Because of the noise and variability inherent in Internet photos and novice annotators, designing this annotation engine was a key challenge; we present a multi-stage set of annotation tasks with quality checks and validation. We demonstrate the use of this database in proof-of-concept applications including surface retexturing and material and image browsing, and discuss future uses. OpenSurfaces is a public resource available at http://opensurfaces.cs.cornell.edu/.

Precise depth calculation is of crucial importance in graphics rendering. Improving precision raises the quality of all downstream graphical techniques that rely on computed depth (e.g., depth buffers, soft and hard shadow maps, screen space ambient occlusion, and 3D stereo projection). In addition, the domain of correctly renderable scenes is expanded by allowing larger far-to-near plane ratios and smaller depth separation between mesh elements. Depth precision is an ongoing problem as visible artifacts continue to plague applications from interactive games to scientific visualizations despite advances in graphics hardware.

In this paper we present and analyze two methods that greatly impact visual quality by automatically improving the precision of depth values calculated in a standard perspective divide rendering system such as OpenGL or DirectX. The methods are easy to implement and compatible with 1/Z depth value calculations. The analysis can be applied to any depth projection based on the method of homogeneous coordinates.

Many tasks in computer vision can be cast as a "label changing" problem, where the goal is to make a semantic change to the appearance of an image or some subject in an image in order to alter the class membership. Although successful task-specific methods have been developed for some label changing applications, to date no general purpose method exists. Motivated by this we propose deep manifold traversal, a method that addresses the problem in its most general form: it first approximates the manifold of natural images then morphs a test image along a traversal path away from a source class and towards a target class while staying near the manifold throughout. The resulting algorithm is surprisingly effective and versatile. It is completely data driven, requiring only an example set of images from the desired source and target domains. We demonstrate deep manifold traversal on highly diverse label changing tasks: changing an individual's appearance (age and hair color), changing the season of an outdoor image, and transforming a city skyline towards nighttime.

We propose a new neural network architecture for solving single-image analogies - the generation of an entire set of stylistically similar images from just a single input image. Solving this problem requires separating image style from content. Our network is a modified variational autoencoder (VAE) that supports supervised training of single-image analogies and in-network evaluation of outputs with a structured similarity objective that captures pixel covariances. On the challenging task of generating a 62-letter font from a single example letter we produce images with 22.4% lower dissimilarity to the ground truth than state-of-the-art.