CognAI is a machine learning, data and AI consultancy in Hong Kong, taking projects from feasibility study and prototype to production software.
A few things we have built and contributed to, spanning research, art and applied machine learning.
Generating music from images, using emotion as the bridge between the visual and auditory domains. Pre-trained image representations drive two music generation architectures, one RNN-based and one Transformer-based. Human and machine evaluations indicate both generate music that carries emotional resonance. By Xiaodong Tan and Mathis Antony, published at the International Conference on Computational Creativity 2020.
An interactive installation: reclining chairs where visitors listen to machine-generated self-improvement advice. Computer vision analyses the participant's facial features, and a text model trained on self-help literature generates a mix of spiritual, motivational and self-help guidance shaped by the detected emotions. By Manolis Perrakis and Mathis Antony, with Xiaodong Tan (text generation) and Cristina Kountiou (textiles). Shown at Ars Electronica 2019 and Sónar Hong Kong 2019.
2nd prize in the Kaggle Diabetic Retinopathy Detection competition. VGG-style convolutional neural networks on large colour retina images, with heavy data augmentation, dynamic resampling for class imbalance, and a per-patient feature blending strategy that combines predictions across both of a patient's eyes.
A generative audio project by Roberto Alonso Trillo and Marek Poliks. We worked with them on the technical implementation, building neural audio synthesis pipelines around descriptor transformers, MelGAN and UnaGAN.
CognAI Limited was established in 2017 in Hong Kong. We provide machine learning consulting services and training, and we value well-tested, maintainable software that makes our clients happy in the long run.
Our solutions span the whole lifecycle, from prototyping and feasibility studies through to production software, including deployment, scaling and monitoring.
We work transparently so you can accurately assess the opportunities and the risks of machine learning solutions before committing to them.
Well-tested and maintainable software over quick hacks: systems your team can understand, operate and extend long after the engagement ends.
Have a project in mind, or just want to find out whether machine learning can help? Drop us a line.
hello@cogn.ai