Repository for Publications and Research Data
News from the ETH Library
Recently Added
Faithful 3D Avatars for Everyone: Prior-Guided Face Reconstruction for High-Quality Novel View Synthesis from Casual Few-Shot Captures
Item type: Doctoral Thesis
Bühler, Marcel; Bühler, Marcel (2025)
Photorealistic digital avatars have many applications in areas ranging from virtual reality and telepresence to digital content creation and human-computer interaction. Today, digital artists create such avatars by capturing a target person in a professional capture dome with hundreds of studio-grade cameras. Reducing the input requirements from hundreds of studio cameras to just a few casually captured smartphone photos has the potential to democratize high-quality face modeling and make avatars widely accessible in previously impossible scenarios.
This thesis studies and develops techniques to reconstruct faces and render high-quality novel views from casual smartphone captures. This problem is inherently under-constrained because the sparse inputs only provide limited and possibly ambiguous information about the underlying geometry. Inspired by how digital artists would use their prior knowledge, this thesis leverages data-driven priors to resolve these ambiguities.
This thesis contributes several insights to the field. First, it establishes that data-driven priors can effectively resolve the fundamental ambiguities in 3D face reconstruction for novel view synthesis from sparse inputs. Second, it demonstrates that these priors can be learned from both real and synthetic datasets, with the latter showing a surprising generalization capability for in-the-wild captures from the real world. Third, the work shows that volumetric representations, when combined with appropriate priors, can capture fine-grained facial details like wrinkles and eyelashes even from minimal input. These insights were developed through three interconnected technical contributions.
We lay the foundation with VariTex, a generative model of neural face textures. The key insight is that integrating a 3D Morphable Face Model and training a generative model in its texture space enables extreme novel poses (beyond 30 degrees) and expressions, even when the training data is strongly biased to frontal and smiling faces. The VariTex model can sample novel identities and render extreme head poses beyond the distribution of the training set. Given a single image of a target person, VariTex reconstructs neural face textures and renders novel views and expressions.
As a drawback, the neural face texture representation in VariTex is only defined on the face surface. Non-surface regions like hair lack consistency across views. Hence, the subsequent model, Preface, replaces the neural-texture-based representation with a Neural Radiance Field (NeRF). As a major challenge, NeRFs require dozens of views for training, which are infeasible to obtain from casual captures. To solve this, Preface proposes a data-driven volumetric prior, which reduces the input requirements to as few as two input images. The prior takes the form of an identity-conditioned NeRF. This conditional NeRF learns the distribution of faces, including non-surface components like glasses and hair. The method uses a sparse landmark-based alignment to create a smooth latent space. This space generalizes to previously unseen subjects, requiring only two input views at inference time.
As a limitation, Preface requires a large dataset of real multi-view captures. Such data are expensive to capture, may exhibit demographic biases, and are subject to strict privacy protection regulations, which hinders general applicability. Fortunately, the third contribution of this thesis, Cafca, shows that a prior can be trained on purely synthetic data. The key insight is that such a purely synthetic prior can bridge the gap to real-world inputs through only minimal finetuning. Cafca first generates a synthetic dataset by combining a geometric face model with assets like hair, clothing, and glasses; and rendering synthetic faces with diverse expressions. Cafca then extends the identity-conditioned prior from Preface with expressions. The synthetic prior in Cafca generalizes to out-of-distribution inputs like in-the-wild smartphone captures and stylized faces.
Together, these contributions enable high-quality face reconstruction and novel view synthesis from casual captures, making photorealistic digital avatars accessible in everyday scenarios previously limited to professional studios.
A Jammer-Resilient 2.87mm² 1.28MS/s 310mW Multi-Antenna Synchronization ASIC in 65nm
Item type: Conference Paper
Arquint, Flurin; Castañeda Fernández, Oscar; Marti, Gian; et al. (2025)
We present the first ASIC implementation of jammerresilient multi-antenna time synchronization. The ASIC implements a recent algorithm that mitigates jamming attacks on synchronization signals using multi-antenna processing. Our design supports synchronization between a single-antenna transmitter and a 16-antenna receiver while mitigating smart jammers with up to two transmit antennas. The fabricated 65 nm ASIC has a core area of 2.87 mm2, consumes a power of 310 mW, and supports a sampling rate of 1.28 mega-samples per second (MS/s)
B cell extracellular vesicles influence melanoma response to immune checkpoint therapy
Item type: Journal Article
Al Hrout, Ala’a; Balayev, Agshin; Cervantes-Gracia, Karla; et al. (2025)
The immune tumor microenvironment is a dynamic ecosystem where B cells play critical roles in modulating immune checkpoint blockade (ICB) therapy responses. While traditionally seen as passive players in tumor immunity, recent evidence suggests that B cells actively influence antitumor responses. This study examines the role of B cells and their extracellular vesicles (EVs) in melanoma responses to ICB. Retrospective meta-analyses reveal increased B cell enrichment in ICB responders’ pretreatment. Functional assays show that B cell depletion impairs T cell–mediated tumor cytotoxicity. EVs from melanoma tumors were analyzed, identifying miR-99a-5p in CD19+ EVs as up-regulated in responders. Silencing miR-99a-5p in B cells reduces T cell antitumor activity, suggesting its role in immune modulation. Mechanistically, miR-99a-5p promotes B cell maturation via class-switch recombination. These findings underscore B cells’ impact on melanoma immunotherapy, offering insights into novel therapeutic strategies targeting B cell–related pathways.
A 14 ns-Latency 9GB/s 0.44mm² 62pJ/b Short-Blocklength LDPC Decoder ASIC in 22FDX
Item type: Conference Paper
Nonaca, Darja; Guichemerre, Jérémy; Wiesmayr, Reinhard; et al. (2025)
Ultra-reliable low latency communication (URLLC) is a key part of 5 G wireless systems. Achieving low latency necessitates codes with short blocklengths for which polar codes with successive cancellation list (SCL) decoding typically outperform message-passing (MP)-based decoding of low-density parity-check (LDPC) codes. However, SCL decoders are known to exhibit high latency and poor area efficiency. In this paper, we propose a new short-blocklength multi-rate binary LDPC code that outperforms the 5G-LDPC code for the same blocklength and is suitable for URLLC applications using fully parallel MP. To demonstrate our code's efficacy, we present a 0.44mm2 GlobalFoundries 22FDX LDPC decoder ASIC which supports three rates and achieves the lowest-in-class decoding latency of 14 ns while reaching an information throughput of 9 Gb/s at 62pJ/b energy efficiency for a rate- 1/2 code with 128-bit blocklength.
Vertical shear resistance of Composite Steel Truss in Concrete (CSTC) beams in fire: Pull-out tests and numerical analyses
Item type: Journal Article
Vigneri, Valentino; Giulieri , Morena; Di Nunzio , Giuseppe; et al. (2025)
The present study introduces a special experimental setup for investigating the anchorage resistance of the diagonal of Composite Steel Truss and Concrete (CSTC) beams at elevated temperatures involved in the vertical shear resistance. The experimental campaign comprised four pull-out tests wherein each specimen was subjected to standard ISO 834–1 fire conditions and a constant load on the diagonal web bar. The observed failure mode in the tests indicated that the pulling resistance is limited by the fracture of the bar occurring near the bottom weld seam, resulting from the heat-degraded mechanical properties of the material. To further expand the experimental findings, finite element (FE) models were developed and validated against the test results via heat transfer and thermomechanical analyses in terms of temperature measurements and steel-concrete slip displacements. The numerical model was subsequently used to establish a benchmark data pool for verifying the suitability of the proposed method to predict the anchorage resistance of the diagonal web bar of CSTC beams in fire. This approach calculates a reference temperature defined at the expected fracture surface via heat transfer analyses. The resulting temperature-dependent yield strength is ultimately considered for the steel component of the vertical shear resistance, which cannot exceed the resistance of the diagonal concrete strut, in accordance with the strut-and-tie model.