AI Power Pause
Kuukauden paras 30 minuuttia tekoälylle!
Tervetuloa mukaan AI Finlandin AI Power Pause -koulutuskahvihetkeen – uuteen kuukausittaiseen webinaarisarjaan, jossa tarjoamme tiiviin puolituntisen tekoälyn ajankohtaisimmista teknisistä teemoista. Toukokuun Power Pausen hostaa kumppanimme Emblica, ja asiantuntijavieraanamme kuulemme Emblican perustaja ja toimistusjohtajaa Teemu Heikkilää sekä Markku Leppälää, Partner and ML Engineer. Englanninkielisen tapahtuman kuvaus:
Synthetic Data for training models out of thin air
When real-world data is scarce or impossible to gather or label, synthetic data generation becomes crucial. This session explores the creation of rich, multimodal datasets from LLM-generated text to sensor streams and control logic derived from physics engines. We dive into the common obstacles such as the ”Sim-to-Real gap,” explaining why models trained in perfect simulations often fail in the messy physical world, and detailing the technical strategies, like domain randomization and noise modeling, to overcome it. Through examples like crane control optimization, natural language data extraction, OCR, and LiDAR-based person detection, you will learn how model synthetic data pipelines can be used to create diverse datasets for multiple modalities and familiarize yourself with technical methods to ensure your model survives the transition from simulation to reality.
Who is this for? Technology and data professionals, developers, project managers, and decision-makers who leverage AI in their work.