Deep Learning for Detecting 3D Underwater Objects
Engineering, IT, Mathematics and Statistics
- This internship is able to cover project costs for domestic students only.
- The Industry Partner has implemented appropriate preparations to comply with Federal and State Government requirements regarding COVID-19 safety.
- If your skillset is aligned with this internship and you are located remotely, please enquire with the Internship Contact to discuss possible arrangements.
ABOUT THE INDUSTRY PARTNER
Veris is the largest provider of surveying, planning and spatial services in Australia. Veris has 20 offices distributed across Australia, and a team of almost 500 professionals. Built through the acquisition and integration of some of the largest and most innovative businesses in the industry, Veris has a unique capability to service large multi-faceted projects in the infrastructure, resource and property sectors. To continue as a market leader Veris is extending its technology capabilities to take advantage of its vast quantity of data in order to provide advanced insights and knowledge to its clients.
WHAT’S IN IT FOR YOU?
- Able to work with cutting-edge 3D capture technology and datasets
- Ability to connect with Australia’s leading spatial professionals
- Real-world applications for industry and defence applications
- Ability to forge a career and contacts within the hydrographic sector
RESEARCH TO BE CONDUCTED
Identifying submerged objects for search and discovery projects, and for safe navigation of shipping, takes significant time within the processing stage of hydrographic data. Each year ports, waterway managers and mariners absorb significant costs ensuring the safe passage of vessels. The current software packages for hydrographic data are only beginning to explore machine learning for detecting underwater objects. Veris is seeking to improve this process by developing deep learning algorithms to identify objects under the water such as logs, rocks, tyres, pipelines, and human-made objects. These objects are captured within 3D point clouds from the sonar equipment, however, require manual review of the data to identify them. Existing software in the market often mistakenly remove these objects as ‘noise’ or erroneous data. Target detection is a key requirement for Defence and Industry project specifications.
The proposed research will need to undertake a review of the current state of machine learning for detecting underwater objects. It will also need to review the current human-powered workflow to identify opportunities for automation. The research will then be concentrated on developing deep learning algorithms to detect both natural and human-made objects along with their key attributes, such as size and other descriptors. Veris has suitable training datasets with the described objects in them.
SKILLS WISH LIST
If you’re a PhD student and meet some or all the below we want to hear from you. We strongly encourage women, indigenous and disadvantaged candidates to apply:
- The ability to work effectively as part of a multi-disciplinary, potentially regionally dispersed team, plus the motivation and discipline to carry out independent research.
- Solid knowledge of computer vision and machine learning, and the ability to understand and develop mathematically founded algorithms and their development in toolkits such as TensorFlow or PyTorch.
- High level computational and programming skills (in Python or C++) to build computer vision/machine learning models and conduct analyses. Python is particularly popular language with existing hydrographic software and offers potentially easier integration.
- Experience or interest in one or more of the following: deep neural networks including recurrent neural networks, zero- or few-shot learning, graph convolutional networks and fine-grained classification.
- A record of science innovation and creativity, including the ability & willingness to incorporate novel ideas and approaches into scientific investigations.
The research outcomes will be a brief report (3-5 pages) of the current best practice for using machine learning to underwater objects. The main research output will be a scalable algorithm that will identify both natural and human-made objects data along with their attributes and uncertainty. Accompanying the algorithms will be a short report on testing, deployment, and the code repository.
The intern will receive $3,000 per month of the internship, usually in the form of stipend payments.
It is expected that the intern will primarily undertake this research project during regular business hours and maintain contact with their academic mentor throughout the internship either through face-to-face or phone meetings as appropriate.
The intern and their academic mentor will have the opportunity to negotiate the project’s scope, milestones and timeline during the project planning stage.
CONNECT WITH APR.INTERN