Fig. 1. Given a single LDR image of an indoor scene, our method automatically predicts HDR lighting (insets, tone-mapped for visualization). Our method learns a direct mapping from image appearance to scene lighting from large amounts of real image data; it does not require any additional scene information, and can even recover light sources that are not visible in the photograph, as shown in these examples. Using our lighting estimates, virtual objects can be realistically relit and composited into photographs.We propose an automatic method to infer high dynamic range illumination from a single, limited field-of-view, low dynamic range photograph of an indoor scene. In contrast to previous work that relies on specialized image capture, user input, and/or simple scene models, we train an end-to-end deep neural network that directly regresses a limited field-of-view photo to HDR illumination, without strong assumptions on scene geometry, material properties, or lighting. We show that this can be accomplished in a three step process: 1) we train a robust lighting classifier to automatically annotate the location of light sources in a large dataset of LDR environment maps, 2) we use these annotations to train a deep neural network that predicts the location of lights in a scene from a single limited field-of-view photo, and 3) we fine-tune this network using a small dataset of HDR environment maps to predict light intensities. This allows us to automatically recover highquality HDR illumination estimates that significantly outperform previous state-of-the-art methods. Consequently, using our illumination estimates for applications like 3D object insertion, produces photo-realistic results that we validate via a perceptual user study.

We present a CNN-based technique to estimate highdynamic range outdoor illumination from a single low dynamic range image. To train the CNN, we leverage a large dataset of outdoor panoramas. We fit a low-dimensional physically-based outdoor illumination model to the skies in these panoramas giving us a compact set of parameters (including sun position, atmospheric conditions, and camera parameters). We extract limited field-of-view images from the panoramas, and train a CNN with this large set of input image-output lighting parameter pairs. Given a test image, this network can be used to infer illumination parameters that can, in turn, be used to reconstruct an outdoor illumination environment map. We demonstrate that our approach allows the recovery of plausible illumination conditions and enables photorealistic virtual object insertion from a single image. An extensive evaluation on both the panorama dataset and captured HDR environment maps shows that our technique significantly outperforms previous solutions to this problem.

An approach for accurately measuring human motion through Markerless Motion Capture (MMC) is presented. The method uses multiple color cameras and combines an accurate and anatomically consistent tracking algorithm with a method for automatically generating subject specific models. The tracking approach employed a Levenberg-Marquardt minimization scheme over an iterative closest point algorithm with six degrees of freedom for each body segment. Anatomical consistency was maintained by enforcing rotational and translational joint range of motion constraints for each specific joint. A subject specific model of the subjects was obtained through an automatic model generation algorithm (Corazza et al. in IEEE Trans. Biomed. Eng., 2009) which combines a space of human shapes (Anguelov et al. in Proceedings SIGGRAPH, 2005) with biomechanically consistent kinematic models and a pose-shape matching algorithm. There were 15 anatomical body segments and 14 joints, each with six degrees of freedom (13 and 12, respectively for the HumanEva II dataset). The overall method is an improvement over (Mündermann et al. in Proceedings of CVPR, 2007) in terms of both accuracy and robustness. Since the method was originally develElectronic supplementary material The online version of this article (http://dx.

Most current single image camera calibration methods rely on specific image features or user input, and cannot be applied to natural images captured in uncontrolled settings. We propose directly inferring camera calibration parameters from a single image using a deep convolutional neural network. This network is trained using automatically generated samples from a large-scale panorama dataset, and considerably outperforms other methods, including recent deep learning-based approaches, in terms of standard L2 error. However, we argue that in many cases it is more important to consider how humans perceive errors in camera estimation. To this end, we conduct a large-scale human perception study where we ask users to judge the realism of 3D objects composited with and without ground truth camera calibration. Based on this study, we develop a new perceptual measure for camera calibration, and demonstrate that our deep calibration network outperforms other methods on this measure. Finally, we demonstrate the use of our calibration network for a number of applications including virtual object insertion, image retrieval and compositing.

No abstract