Pushing AI frontiers, globally.
Leading with deep domain expertise.
At K2 Research, we push the boundaries of AI and machine learning. Our researchers collaborate with the world's best academic institutions and combine a varied set of deep domain skills.
K2 Research is paramount to K2's mission in general. Expertise fuels excellence - our experts operate across our AI execution hub, coming up with solutions for the most challenging of problems.
WHAT WE DO
Our Projects
EDUCATED AT
WHO WE ARE
Our People
SCIENTIFIC CONTRIBUTIONS
Selected Research
01
Cechnicka et al. present a novel method to enhance robustness in histopathological image segmentation, addressing challenges of limited datasets and imbalanced object representation. Their approach utilizes diffusion models to enrich datasets with realistic examples, improving performance and interpretability. Validation on two datasets demonstrates its efficacy for clinical applications.
02
A. Obuchowski et al. introduce an innovative approach to intent recognition, merging transformer architecture with capsule networks. Their method outperforms traditional capsule-NLU networks, achieving state-of-the-art results on key datasets like ATIS, AskUbuntu, and WebApp. This advancement holds significant promise for enhancing intelligent conversation design within NLU systems.
03
Renal tumor malignancy classification profoundly impacts urological decisions, yet misdiagnoses lead to unnecessary nephrectomies. A machine-aided diagnosis system, presented by A. Obuchowski et al., predicts malignancy post radiological diagnosis to mitigate false-positives. Achieving an F1-score of 0.84, their method incorporates colorization-based preprocessing for improved knowledge transfer, enhancing accuracy by up to 1.8pp.