A Perceptual Measure for Deep Single Image Camera and Lens Calibration.

Image editing and compositing have become ubiquitous in entertainment, from digital art to AR and VR experiences. To produce beautiful composites, the camera needs to be geometrically calibrated, which can be tedious and requires a physical calibration target. In place of the traditional multi-image calibration process, we propose to infer the camera calibration parameters such as pitch, roll, field of view, and lens distortion directly from a single image using a deep convolutional neural network. We train this network using automatically generated samples from a large-scale panorama dataset, yielding competitive accuracy in terms of standard l2 error. However, we argue that minimizing such standard error metrics might not be optimal for many applications. In this work, we investigate human sensitivity to inaccuracies in geometric camera calibration. To this end, we conduct a large-scale human perception study where we ask participants to judge the realism of 3D objects composited with correct and biased camera calibration parameters. Based on this study, we develop a new perceptual measure for camera calibration and demonstrate that our deep calibration network outperforms previous single-image based calibration methods both on standard metrics as well as on this novel perceptual measure. Finally, we demonstrate the use of our calibration network for several applications, including virtual object insertion, image retrieval, and compositing.

A high-Z inorganic scintillator-based detector for time-resolved in vivo dosimetry during brachytherapy.

PURPOSE: High-dose rate (HDR) and pulsed-dose rate (PDR) brachytherapy would benefit from an independent treatment verification system to monitor treatment delivery and to detect errors in real time. This paper characterizes and provides an uncertainty budget for a detector based on a fiber-coupled high-Z inorganic scintillator capable of performing time-resolved in vivo dosimetry during HDR and PDR brachytherapy. METHOD: The detector was composed of a detector probe and an optical reader. The detector probe consisted of either a 0.5 × 0.4 × 0.4 mm3 (HDR) or a 1.0 × 0.4 × 0.4 mm3 (PDR) cuboid ZnSe:O crystal glued onto an optical-fiber cable. The outer diameter of the detector probes was 1 mm, and fit inside standard brachytherapy catheters. The signal from the detector probe was read out at 20 Hz by a photodiode and a data acquisition device inside the optical reader. In order to construct an uncertainty budget for the detector, six characteristics were determined: (1) temperature dependence of the detector probe, (2) energy dependence as a function of the probe-to-source position in 2D (determined with 2 mm resolution using a robotic arm), (3) the signal-to-noise ratio (SNR), (4) short-term stability over 8 h, and (5) long-term stability of three optical readers and four probes used for in vivo monitoring in HDR and PDR treatments over 21 months (196 treatments and 189 detector calibrations, and (6) dose-rate dependence. RESULTS: The total uncertainty of the detector at a 20 mm probe-to-source distance was < 5.1% and < 5.8% for the HDR and PDR versions, respectively. Regarding the above characteristics, (1) the sensitivity of the detector decreased by an average of 1.4%/°C for detector probe temperatures varying from 22 to 37°C; (2) the energy dependence of the detector was nonlinear and depended on both probe-to-source distance and the angle between the probe and the brachytherapy source; (3) the median SNRs were 187 and 34 at a 20 mm probe-to-source distance for the HDR and PDR versions, respectively (corresponding median source activities of 4.8 and 0.56 Ci, respectively); (4) the detector response varied by 0.6% in 11 identical irradiations over 8 h; (5) the sensitivity of the four detector probes decreased systematically by 0-1.2%/100 Gy of dose delivered to the probes, and random fluctuations of 4.8% in the sensitivity were observed for the three probes used in PDR and 1.9% for the probe used in HDR; and (6) the detector response was linear with dose rate. CONCLUSION: ZnSe:O detectors can be used effectively for in vivo dosimetry and with high accuracy for HDR and PDR brachytherapy applications.