Student projects and final year projects at the Chair of Media Technology

Can generative AI transform a story described through text, images, sketches, or other modalities into a consistent and editable 3D storyboard?

Recent advances in generative AI enable the creation of impressive images and videos from textual prompts. However, creating a coherent visual story across multiple shots remains challenging. Characters may change appearance, objects can move unexpectedly, and spatial relationships may be lost between shots.

The recent StoryBlender: Inter-Shot Consistent and Editable 3D Storyboard with Spatial-temporal Dynamics approach addresses this problem by generating structured 3D storyboards rather than independent 2D images. Its pipeline combines semantic-spatial grounding, persistent assets in a common 3D space, spatial-temporal dynamics, and multi-agent planning with verification.

In this thesis, you will investigate how these ideas can be extended toward interactive multimodal story generation.

Your Task

The goal is to design and prototype a system that translates a multimodal story description into a sequence of coherent 3D scenes or shots.

Depending on the focus of the thesis, the system could accept combinations of:

• Natural-language descriptions or scripts
• Reference images for characters, objects, or environments
• Sketches or spatial layouts
• Existing 3D models or scenes
• Potentially speech or other interaction modalities

The system should interpret these inputs, construct an internal representation of the story, and generate a sequence of consistent and editable 3D shots.

Possible Research Directions

The thesis could investigate one or more of the following:

Agentic generation

Can specialized AI agents take responsibility for tasks such as script interpretation, scene composition, asset selection, camera placement, animation, and verification?

Interactive editing

How can users modify generated results using natural language, for example:

“Move the camera closer to the character.”

“Keep the same room, but make the next shot an over-the-shoulder shot.”

“Replace the robot with this reference model in all shots.”

A key advantage of a 3D representation is that cameras and scene assets can subsequently be changed directly rather than requiring the entire image to be regenerated, which is also one of the motivations behind StoryBlender.

Prerequisites

How can Virtual Reality make the creation of stop-motion animations faster, more intuitive, and more accessible?

Stop-motion animation is a creative but labor-intensive process: objects must be positioned manually, photographed frame by frame, and repeatedly adjusted to achieve the desired motion. Virtual Reality offers an opportunity to rethink this workflow by allowing users to create and manipulate animations directly in an immersive 3D environment.

Your Task

In this master’s thesis, you will design and develop a VR application for Meta Quest that supports the creation of stop-motion animations.

The thesis will involve:

• Designing an intuitive VR interaction concept for creating stop-motion animations
• Developing a VR prototype for Meta Quest
• Choosing and evaluating an appropriate development approach, for example:
  • Unity / OpenXR for a native VR application, or
  • Immersive Web SDK (IWSDK) for a WebXR and browser-based 3D application
• Implementing functionality for positioning and manipulating objects, defining animation frames, previewing motion, and supporting repetitive animation tasks
• Exploring approaches for partially automating the stop-motion workflow
• Conducting a user study to evaluate the proposed approach
• Analyzing usability, efficiency, learnability, and overall user experience

Possible Research Questions

• Can VR reduce the effort required to create stop-motion animations?
• Which parts of the stop-motion workflow can be effectively automated?
• How accurately and efficiently can users define object motion in VR?
• How does a VR-based workflow compare with conventional stop-motion animation in terms of task completion time, usability, and user experience?
• What are the advantages and limitations of a native VR implementation versus a WebXR-based approach for this type of application?

Prerequisites

This work falls within the scope of indoor localization, more precisely the fingerprinting-based indoor localization. 

Your task will be to investigate the potential and outcomes of opting for a multi-layer reference map during the "Offline phase", which represents a potential improvement idea that has never been adopted or tested in current state-of-the-art fingerprinting schemes.

The aim here is to achieve a better trade-off between both performance and costs, yielding a better localization method. 

Prerequisites

The Chair of Media Technology at the Technical University of Munich (TUM) is offering a Master's/Bachelor's Thesis (or guided research) opportunity in the context of model-based deep learning for indoor localization, combining classical Kalman filtering theory with modern neural state estimation.

Indoor localization using WiFi RSSI and inertial sensors (IMU) is a sequential state-estimation problem naturally suited to Kalman filtering, where the IMU drives the motion model and WiFi RSSI provides the correction. However, the relationship between RSSI and position is highly nonlinear and building-specific, motivating a hybrid model- and data-driven approach. KalmanNet replaces the analytically-derived Kalman gain with a GRU-based recurrent network learned from data, retaining the interpretability and low-data efficiency of the classical filter while implicitly learning complex, unmodeled dynamics.

The thesis investigates KalmanNet on the IMUWiFine dataset (220 per-AP RSSI values, 9-channel IMU, fine-grained ground-truth trajectories across a multi-floor building) along a complexity ladder: from a fully active system with complete per-access-point observability, through intermediate degradation stages (AP reduction, sectorization), to a fully passive distributed antenna system (DAS), where only a single aggregated RSSI value is available per time step.

Selection and Reproduction of Baseline

As KalmanNet has not previously been applied to this dataset, the first part of the thesis involves reproducing a KalmanNet baseline and validating it against classical filters (EKF, UKF) and the existing LSTM-based end-to-end baseline. The implementation should support:

• Per-AP RSSI and per-antenna observation vectors

• IMU-driven state-evolution modeling

• GRU-based Kalman gain estimation, trained end-to-end

Degradation Analysis and Passive DAS Reformulation

Building upon the baseline above, the second part focuses on systematically degrading observability and identifying what must change for KalmanNet to remain viable as the system moves toward a passive

DAS. This involves:

• Reducing and sectorizing the AP/antenna observation set

• Investigating the non-invertibility and dimensionality collapse of aggregated RSSI observations

• Synthesizing aggregated RSSI from existing per-AP data to emulate the passive DAS setting

• Increasing reliance on IMU-based state evolution to compensate for degraded observability

Prerequisites

Large-scale diffusion models and, similarly, flow models are strong natural image priors [1,2]. This capability is enabled by both architectural advancements, such as Transformers [3], and internet-scale training.

This capability of the diffusion models enabled perceptual image restoration where non-generative or small-scale models failed to produce realistic restorations [4]. However, these methods rely on external feature extractors to condition the diffusion backbones on the degraded inputs. We showed that this is unnecessary, as the backbone itself exhibits a degree of robustness to degradation [5]. However, this requires processing both the degraded input and the diffused image simultaneously, resulting in high computational complexity.

This project will focus on improving the performance of this family of methods.

[1] Rombach, Robin, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. "High-resolution image synthesis with latent diffusion models." In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 10684-10695. 2022.

[2] Liu, Xingchao, Chengyue Gong, and Qiang Liu. "Flow straight and fast: Learning to generate and transfer data with rectified flow." arXiv preprint arXiv:2209.03003 (2022).

[3] Peebles, William, and Saining Xie. "Scalable diffusion models with transformers." In Proceedings of the IEEE/CVF international conference on computer vision, pp. 4195-4205. 2023.

[4] Li, Xin, Yulin Ren, Xin Jin, Cuiling Lan, Xingrui Wang, Wenjun Zeng, Xinchao Wang, and Zhibo Chen. "Diffusion models for image restoration and enhancement: A comprehensive survey." International Journal of Computer Vision 133, no. 11 (2025): 8078-8108.

[5] Eteke, Cem, Alexander Griessel, Wolfgang Kellerer, and Eckehard Steinbach. "BIR-Adapter: A parameter-efficient diffusion adapter for blind image restoration." Pattern Recognition (2026): 113824.

Prerequisites

In today's digital era, image and video content dominate online traffic, accounting for the majority of global data transmission [1]. Efficient compression is therefore essential for delivering high-quality content under limited bandwidth and storage constraints. Recently, Deep Neural Networks (DNNs) have emerged as a powerful alternative to traditional compression methods, leveraging nonlinear representations to improve compression efficiency and visual quality [1].

However, beyond compression performance, practical deployment requires careful consideration of computational cost and energy consumption, especially on resource-constrained devices such as smartphones. Importantly, neither the number of parameters nor the number of operations in a DNN directly reflects its true energy consumption. Models with fewer parameters or operations may still consume more energy due to hardware-specific factors such as memory access patterns [2].

This project aims to systematically measure and model the runtime and energy consumption of DNNs across heterogeneous devices, including desktop GPUs, laptops, and/or smartphones. The objectives include refining and extending an existing energy measurement setup to ensure reliability and reproducibility, analyzing the relationship between model structure, runtime, and energy consumption, and designing predictive models for runtime and energy estimation across devices.

Existing approaches in the literature include device-specific [3] and device-adaptive [4] runtime estimation methods. Students are encouraged to explore and extend these approaches and to propose novel runtime and energy modeling techniques for the generalization across devices.

[1] JS Gomes, M. Grellert, FLL Ramos and S. Bampi, "End-to-End Neural Video Compression: A Review," IEEE Open Journal of Circuits and Systems, vol. 6, pp. 120-134, 2025.

[2] X. Yang, J. Kwon, Y. Li and Y. Chen, "Designing Energy-Efficient Convolutional Neural Networks using Energy-Aware Pruning," CVPR, 2017.

[3] LL Zhang, S. Han, J. Wei, N. Zheng, T. Cao, Y. Yang and Y. Liu, "NN-Meter: Towards Accurate Latency Prediction of Deep-Learning Model Inference on Diverse Edge Devices," MobiSys, pp. 81-93, 2021.

[4] C. Feng, LL Zhang, Y. Liu, J. Xu, C. Zhang, Z. Wang, T. Cao, M. Yang and H. Tan, "LitePred: Transferable and Scalable Latency Prediction for Hardware-Aware Neural Architecture Search," NSDI, 2024.

Prerequisites