Internship: Evaluation and Integration of Deep Learning-Based Object Detection and Tracking Algorithms (F/M)

We are looking for an intern to evaluate and integrate deep learning-based object detection and tracking algorithms (F/M) for a duration of 6 months starting in Q1 2027 within the Land & Sea division at the Angoulême site. Context: Deep learning-based detection and tracking algorithms are evolving rapidly, and their announced performance transfers poorly to new sensors or new conditions. Our team develops real-time vision solutions and has a modular library of processing pipelines (C++). There is a lack of a reliable method to compare approaches and choose which ones to integrate. Objective: Set up a reproducible evaluation platform, then integrate the most relevant solutions into the library. Missions: - Review of the existing system. Analyze the current detection and tracking module, diagnose its generalization limits on other sensors, and report on them. - Data and evaluation protocol. Build infrared and visible datasets, with splitting by sequence, cross-dataset tests, and standardized metrics (mAP, HOTA, IDF1, execution speed). - State-of-the-art benchmark. Compare recent detectors and trackers, with training or fine-tuning, data augmentation, and reproducible results. - Integration. Integrate the selected solutions into the C++ library and carry out demonstrations. - Documentation. Write a decision guide per use case and the technical documentation. Profile: - Computer science student specializing in artificial intelligence / computer vision / software engineering with: - Python and PyTorch, and solid foundations in deep learning and model evaluation; - C++ and CMake; - Rigor, autonomy, and well-argued critical thinking. Find more English Speaking Jobs in France on Arbeitnow