Our software helps asset managers replace manual inspection with robotic inspection of virtually anything - bridges, buildings, pipes, cables, telegraph poles - even animals! Our software generates synthetic images that helps companies access machine learning to improve predictive asset maintenance.
Our machine learning software creates data (millions of photographs) where there are none. We help you skip the costly and time-consuming labelling process and create machine learning models that help you detect things that are humans can’t. Our unique simulator can generate millions of ‘synthetic’ images and all the labelling & segmentation data to enable you to train the model. For the techies: • We use capture, CAD & 3D Modelling to simulate your assets, equipment and environment and create photo realistic simulations capturing variants, defects and different environments from every angle. Unlimited data sets optimize machine learning models to achieve highly accurate detection results. • Key highlights of what our software can do: o Project and Image Capture Assessment o 3D Modelling, Photorealistic Textures o Simulation and Randomization, Image Altering o Automated Labelling, Classification and Training
Our simulator can create data (photos and labelling) where there is none – for example generating randomizable defects in assets and equipment, to allow for training recognition models in situations where there aren’t enough real photos to train for accuracy. As the content is synthetic, we have complete control over lighting, camera angles, environment and defects, and can generate pixel perfect labelling data at the same time as the photos. We create data where there is none available, and we skip the expensive and labour intensive labelling step. Our long term plans include standardized architecture for various types of machine learning models, allowing us to run a simulation and generate data, automatically train a model, tweak the simulation and train a second one, then compare accuracy on both and keep the best one. This level of automation can help us generate highly accurate detection models and commoditize computer vision. Image generation is also only the first step, we plan to expand to LIDAR, point cloud and thermal content, audio and text generation and data construction for NLP. We are focussed on reducing the cost and effort involved in training and experimenting with machine learning models in order to make these technologies more affordable and more flexible for industry. Benefits of predictive maintenance: • you receive real time alerts when defects are detected • you get high res image sets without needing to source real photos • you get ‘pixel-perfect’ labelling saving you time and money • we take the effort out of training your machines to learn
We won the ‘Most Innovative Startup’ Pitch Award at the Asset Maintenance Conference in 2019. We’ve had deep interest from corporate Australia after launching just 4 months ago. Companies like KPMG, Downer, AusNet, Fortescue, Deloittes and others. We are on the cusp of working with AusNet for a trial/POC to start in Feb 2020.