Internship | Military image transformation with reinforcement learning-based diffusion

About this internship

Recent advances in generative AI have enabled text-guided image editing, but most existing methods remain limited to appearance changes such as style or texture. In contrast, many defense and security applications require precise geometric control over objects in a scene – for example translating, rotating, or resizing vehicles, sensors, or equipment based on high-level instructions. Recent research introduces a new paradigm by combining diffusion models with reinforcement learning (RL) optimization algorithms to achieve object-level geometric transformations driven by natural language, without relying on costly paired supervision.

Internship | Military image transformation with reinforcement learning-based diffusion

This thesis project builds directly on these ideas and explores how RL-guided diffusion models can be applied and extended to military-relevant imagery and scenarios. The focus is on text-instructed manipulation of objects such as ground vehicles, aircraft, or infrastructure elements in complex scenes, enabling controllable synthetic data generation and scene editing. Such capabilities are highly valuable for simulation, training, and robustness testing of military computer vision systems, where real data is often scarce, sensitive, or operationally constrained.

Possible research directions include:

• Adapting RL-based diffusion frameworks to military scenes with clutter, camouflage, and occlusions.
• Designing object-centric spatial reward functions tailored to defense use cases (e.g., tactical repositioning, formation changes, or sensor line-of-sight constraints).
• Exploring robustness of text-instructed geometric transformations under domain shifts (civilian → military environments).
• Evaluating how RL-guided geometric edits can improve downstream tasks such as object detection, tracking, or scene understanding.
• Investigating extensions to non-RGB modalities (e.g., infrared or thermal imagery) where geometric correctness is critical.

The goal of this thesis is to research and prototype reinforcement learning–based diffusion methods for text-guided object-level geometric transformations. You will study how spatial manipulation can be formulated as a sequential decision-making problem and how RL rewards can be used to align geometric changes with linguistic intent.

You will work on state-of-the-art diffusion and RL techniques with our high-end GPU cluster, implement and adapt existing methods, and evaluate them on military-relevant datasets or scenarios. A key challenge is bridging the gap between general-purpose generative models – typically trained on civilian imagery – and the specific requirements of defense applications. You are expected to critically analyze limitations of current approaches and propose improvements or extensions that increase spatial accuracy, interpretability, and applicability in operational contexts.

You will perform this assignment within TNO’s Intelligent Imaging department. Intelligent Imaging is a dynamic and interdisciplinary team of approximately 60 experts working on advanced computer vision and AI solutions, ranging from medical imaging to defense and security applications. The department has strong expertise in deep learning, generative models, and applied AI research, and offers a stimulating environment for high-impact thesis work.

What we expect from you

We are looking for a motivated master’s student with a strong interest in generative AI, reinforcement learning, and computer vision, and an affinity with defense-related applications. This position is well suited for a student who enjoys combining theoretical concepts with hands-on experimentation.

Requirements:

• You are in the final phase of a master’s degree in Artificial Intelligence, Computer Science, Electrical Engineering, Applied Mathematics, Physics, or a related field.
• Solid background in machine learning and deep learning.
• Experience with computer vision and Python-based research code.
• Familiarity with diffusion models and/or reinforcement learning is a strong plus.
• Interest in text-conditioned models, scene understanding, or synthetic data generation.

What you'll get in return

• A highly professional, innovative internship environment, within a team of top experts.
• A suitable internship allowance (615 euro for wo-, hbo- and mbo-students, for a full-time internship).
• Possibility of eight hours of free leave per internship month (for a full-time internship).
• A free membership of Jong TNO, where you can meet other TNO professionals and join several activities, such as sports activities, (work-related) courses or the yearly ski-trip.
• Use of a laptop.
• An allowance for travel expenses in case you don’t receive an OV-card.

TNO as an employer

At TNO, we innovate for a healthier, safer and more sustainable life. And for a strong economy. We find each other in wonder and ingenuity. We are driven to push boundaries. There is all the space and support for your talent and ambition. You work with people who will challenge you: who inspire you and want to learn from you. Our state-of-the-art facilities are there to realize your vision. What you do at TNO matters: impact makes the difference. Because with every innovation you contribute to tomorrow’s world.

Security clearance requirement

For this internship vacancy it is required that the AIVD issues a security clearance (VGB) after conducting a security screening. Take into account that this process may take about 8 weeks. If you have been abroad for more than 6 consecutive months, or if you do not have the Dutch nationality, it may take longer.