Rotoscoping of facial features is often an integral part of Visual Effects post-production, where the parametric contours created by artists need to be highly detailed, consist of multiple interacting components, and involve significant manual supervision. Yet those assets are usually dis-carded after compositing and hardly reused. In this paper, we present the first methodology to learn from these assets. With only a few manually rotoscoped shots, we identify and extract semantically consistent and task specific landmark points and re-vectorize the roto shapes based on these land-marks. We then train two separate models - one to predict landmarks based on a rough crop of the face region, and the other to predict the roto shapes using only the inferred landmarks from the first model. In preliminary production testing, 26% of shots rotoscoped using our tool were able to be used with no adjustment, and another 47% were able to be used with minor adjustments. This represents a significant time savings for the studio, as artists are able to rotoscope almost 73% of their shots with no manual roto-scoping and some spline adjustment. This paper presents a novel application of machine learning to professional interactive rotoscoping, a methodology to convert unstructured roto shapes into a self-annotated, trainable dataset that can be harnessed to make accurate predictions on future shots of a similar object, and a limited dataset of rotoscoped multi-shape fine feature systems from a real film production.

We study an estimation problem where a fusion center estimates a random parameter by using a partially connected network of sensor nodes. The process involves two stages. In the collaboration stage, the sensor nodes share their observations with their neighbors. In the estimation stage, all the sensor nodes form a coherent beam to the fusion center using the analog amplify-and-forward procedure. In the previous work of Kar and Varshney (2013) on this topic, a control center determines the optimum collaboration strategy that is sent to the sensor nodes prior to starting the two-stage procedure. In this paper, we develop a new framework where the collaboration strategies are computed in a decentralized manner using minimal communication with the control center. This makes the sensor network more energy efficient and reduces control channel communication requirements.

In the context of distributed estimation we consider the problem of sensor collaboration, which refers to the act of sharing measurements with neighboring sensors prior to transmission to a fusion center. While incorporating the cost of sensor collaboration, we aim to find optimal sparse collaboration topologies subject to a certain information or energy constraint. To achieve this goal, we present a tractable optimization framework and propose efficient methods to solve the formulated sensor collaboration problems. The effectiveness of our approach is demonstrated by numerical examples.

A power-constrained sensor network that consists of multiple sensor nodes and a fusion center (FC) is considered, where the goal is to estimate a random parameter of interest. In contrast to the distributed framework, the sensor nodes may be partially connected, where individual nodes can update their observations by (linearly) combining observations from other adjacent nodes. The updated observations are communicated to the FC by transmitting through a coherent multiple access channel. The optimal collaborative strategy is obtained by minimizing the expected mean-square error subject to power constraints at the sensor nodes. Each sensor can utilize its available power for both collaboration with other nodes and transmission to the FC. Two kinds of constraints, namely the cumulative and individual power constraints, are considered. The effects due to imperfect information about observation and channel gains are also investigated. The resulting performance improvement is illustrated analytically through the example of a homogeneous network with equicorrelated parameters. Assuming random geometric graph topology for collaboration, numerical results demonstrate a significant reduction in distortion even for a moderately connected network, particularly in the low local signal-to-noise ratio regime.

We consider a power-constrained sensor network, consisting of multiple sensor nodes and a fusion center (FC), that is deployed for the purpose of estimating a common random parameter of interest. In contrast to the distributed framework, the sensor nodes are allowed to update their individual observations by (linearly) combining observations from neighboring nodes. The updated observations are communicated to the FC using an analog amplify-and-forward modulation scheme and through a coherent multiple access channel. The optimal collaborative strategy is obtained by minimizing the cumulative transmission power subject to a maximum distortion constraint. For the distributed scenario (i.e., with no observation sharing), the solution reduces to the power-allocation problem considered by Xiao et. al.. Collaboration among neighbors significantly improves power efficiency of the network in the low local-SNR regime, as demonstrated through an insightful example and numerical simulations.