<p>3D character modeling is a crucial process of asset creation in the entertainment industry, particularly for animation and games. A fully automated pipeline via sketch-based 3D modeling (SBM) is an emerging possibility, but development is stalled by unrefined outputs and a lack of character-centered tools. This thesis proposes an improved method for constructing 3D character models with minimal user input, using only two sketch inputs  i.e., a front and side unshaded sketch. The system implements a deep convolutional neural network (CNN), a type of deep learning algorithm extending from artificial intelligence (AI), to process the input sketch and generate multi-view depth, normal and confidence maps that offer more information about the 3D surface. These are then fused into a 3D point cloud, which is a type of object representation for 3D space. This point cloud is converted into a 3D mesh via an occupancy network, involving another CNN, for a more precise 3D representation. This reconstruction step contends with non-deep learning approaches such as  Poisson reconstruction. The proposed system is evaluated for character generation on standardized quantitative metrics (i.e., Chamfer Distance [CD], Earth Mover’s Distance [EMD], F-score and Intersection of Union [IoU]), and compared to the base framework trained on the same character sketch and model database. This implementation offers a  significant improvement in the accuracy of vertex positions for the reconstructed character models. </p>

Three-finger toxins (3FTX) are a group of peptides that affect multiple receptor types. One group of proteins affected by 3FTX are nicotinic acetylcholine receptors (nAChR). Structural information on how neurotoxins interact with nAChR is limited and is confined to a small group of neurotoxins. Therefore, in silico methods are valuable in understanding the interactions between 3FTX and different nAChR subtypes, but there are no established protocols to model 3FTX–nAChR interactions. We followed a homology modeling and protein docking protocol to address this issue and tested its success on three different systems. First, neurotoxin peptides co-crystallized with acetylcholine binding protein (AChBP) were re-docked to assess whether Rosetta protein–protein docking can reproduce the native poses. Second, experimental data on peptide binding to AChBP was used to test whether the docking protocol can qualitatively distinguish AChBP-binders from non-binders. Finally, we docked eight peptides with known α7 and muscle-type nAChR binding properties to test whether the protocol can explain the differential activities of the peptides at the two receptor subtypes. Overall, the docking protocol predicted the qualitative and some specific aspects of 3FTX binding to nAChR with reasonable success and shed light on unknown aspects of 3FTX binding to different receptor subtypes.

Magnetic resonance elastography (MRE) is an imaging technique that allows for non-invasive access to the physical properties of body tissues. MRE has great potential, but it is difficult to conduct research due to the time-consuming estimation of stiffness maps, which could be speeded up by using neural network. However, there is not enough real data to train one, thus, synthetic data is needed. To create synthetic data three techniques of simulating tissue displacement due to wave propagation was explored, including solving differential equations for a system of coupled harmonic oscillators (CHO method) and using two different functions from the k-Wave toolbox. Each of the three methods demonstrated the ability to replicate the displacement pattern in a phantom with a simple structure. The CHO method and \texttt{kspaceFirstOrder} function of the k-Wave toolbox showed the best performance when simulating displacement in a 2D brain slice. The models are not very accurate, but capture general features of displacement in a brain and hold potential for future improvement. / Magnetresonans-elastografi (MRE) är en avbildningsteknik som möjliggör icke-invasiv åtkomst till de fysiska egenskaperna hos olika vävnader. MRE har stor potential, men forskning inom ämnet försvåras på grund av den tidskrävande beräkningen av elasticitetskartorna, vilket kan påskyndas med hjälp av ett neuralt nätverk. Dock finns det inte tillräckligt med experimentiell data för att träna ett sådant nätverk, och därför behövs syntetisk data. För att skapa sådan syntetisk MRE-data utforskades tre tekniker för att simulera vågrörelser i hjärnvävnad; dessa tekniker inkluderar lösning av differentialekvationer för ett system av kopplade harmoniska oscillatorer (CHO-metoden) och användning av två olika funktioner från det Matlab-baserade programmet k-Wave. Var och en av de tre metoderna visade potential att återskapa vågsmönstret i en enkel strukturerad fantom. CHO-metoden och funktionen kspaceFirstOrder från k-Wave visade bäst prestanda vid simulering av vågrörelser i ett 2D-segment av hjärnan. Modellerna visade sig inte vara särskilt precisa, men fångar allmänna, kvalitativa, egenskaper av vågrörelser i hjärnan och uppvisar potential för framtida förbättring.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.