D

Deep Learning

Semester project for AIN "Deep Learning and Big Data" (SS 2026) — implementing and comparing two deep learning approaches on the same dataset across two related tasks.This project is a hands-on comparative study of deep learning methods. Rather than building a single "best" model, we implement two fundamentally different models and evaluate them on two related tasks derived from the same dataset. Model 1 is built and trained from scratch — every design choice (architecture, loss, optimizer, hyperparameters) is made and justified by the team. Model 2 follows an alternative paradigm: transfer learning, a pretrained backbone, or a strong classical baseline. By holding the data constant and varying the approach, we isolate the impact of modeling choices and report quantitative + qualitative trade-offs in the final paper and presentation.