Project collaboration within computer vision
Eltronic Group
At Dynatest we are looking for students who seek an exciting and innovative
project collaboration
Do you want to develop your knowledge and skills within computer vision? then read more about your options for a project collaboration with Dynatest!
Dynatest is the pavement industry’s global partner on pavement evaluation solutions. We develop, manufacture, and service equipment and software that defines the industry standard.
Below you can read about our four different projects within computer vision.
Project 1: Computer vision for object detection of pavement distresses
Project description
Our roads are the backbone of connectivity, enabling the flow of commerce, facilitating travel between home and work, and also provide the grounds for many hobbies world wide.
This project focuses on developing an object detection model for detecting multiple pavement distresses based on state-of-the-art methods.
There is a large collection of publicly available datasets on sites such as Kaggle, GitHub, and similar. These datasets can offer a good starting point for the practical work on evaluation of the model(s); however, the student is also free to collect their own dataset.
The proposed project approach is:
- Literature research for identification of appropriate methodologies; models, datasets, evaluation metrics, etc.
- Identify publicly available datasets.
- Collect your own dataset for validation purposes.
- Implement chosen model(s).
- Evaluate model(s).
- Perform data-centric optimization; analyze images and results to implement appropriate preprocessing.
- Evaluate model(s) again.
- Literature research for identification of appropriate methodologies; models, datasets, evaluation metrics, etc.
- Identify publicly available datasets.
- Collect your own dataset for validation purposes.
- Implement chosen model(s).
- Evaluate model(s).
- Perform data-centric optimization; analyze images and results to implement appropriate preprocessing.
- Evaluate model(s) again.
- Literature research for identification of appropriate methodologies; models, datasets, evaluation metrics, etc.
- Identify appropriate datasets.
- Choose a publicly available dataset.
- Collect your own dataset.
- Implement chosen methods.
- Evaluate the implemented method.
- Create a proof-of-concept for linking the data with a GIS system.
- Literature research for identification of appropriate methodologies; models, datasets, evaluation metrics, etc.
- Identify publicly available datasets.
- Collect your own dataset for validation purposes.
- Implement chosen model(s).
- Evaluate model(s).
- Perform data-centric optimization; analyze images and results to implement appropriate preprocessing.
- Evaluate model(s) again.
Project 2: Computer vision for pothole detection
Project description
Potholes pose significant risks to drivers, leading to accidents, vehicle damage, and road hazards. Many road owners are interested in knowing if there are potholes present on their road network and where they are located.
This project focuses on developing a computer vision model based on state-of-the-art methods for a pothole detection system.
There is a large collection of publicly available datasets on sites such as Kaggle, GitHub, and similar. These datasets can offer a good starting point for the practical work on evaluation of the model(s); however, the student is also free to collect their own dataset.
The proposed project approach is:
Project description
In the ever-evolving landscape of transportation engineering, the ability to detect and map the environment surrounding our roadways is becoming increasingly interesting.
Spatial relationships between the road and its adjacent features, such as trees, traffic signs, and other landmarks, holds one of the keys to enhancing safety, optimizing road planning, and road maintenance.
This project focuses on developing a computer vision pipeline for creating a 3D reconstruction of the environment of the inspection vehicle to provide depth information such as distance between the road and trees, signs, or other relevant features. The acquired data can be mapped into GIS systems to create rich multidimensional representations of the road environment.
The proposed project approach is:
- Project 4: Computer vision for crack segmentation
Project description
One of the most common and concerning issues faced by road authorities globally is the occurrence of pavement cracks. These pavement cracks can be caused by non-visible factors or themselves lead to accelerated road deterioration if left unattended.
This project focuses on developing an object segmentation model based on state-of-the-art methods for segmentation of pavement cracks.
There is a large collection of publicly available datasets on sites such as Kaggle, GitHub, and similar. These datasets can offer a good starting point for the practical work on evaluation of the model(s); however, the student is also free to collect their own dataset.
The proposed project approach is:
The student is free to choose between traditional computer vision approaches, deep learning based approaches, or something entirely different. We expect that the project will be based on state-of-the-art methods.
Project supervisors
You are offered professional guidance from Director SW & NPD in Dynatest A/S.
We offer
A very existing collaboration with a fast-growing, innovative, and well-reputed company that is the market leader in the field of pavement evaluation.
Dynatest
As apart of Eltronic Group, Dynatest is an engineering company that develops and manufactures high-tech pavement testing equipment, and we export our innovative and fascinating technical equipment across the globe.
We primarily produce precision equipment for non-destructive assessment of roads and pavement surfaces incl. their load capacity. Our clients are road administrators, advisors, and research institutes. We operate in every corner of the world from our offices and production facilities in Ballerup, Denmark, and Gainesville, Florida USA.
Questions and application
For more information about project collaboration, you are very welcome to contact Director SW & NPS, Emil Ancker: +45 20 22 30 98 / [email protected].
Starting date: By appointment.
We look forward hearing from you!
Department: 50230 SW & NPD
Deadline: 31 January 2024
Location: HQ / Tempovej 27, 2750, Ballerup, Denmark
Opslaget er indhentet automatisk fra virksomhedens jobsider og vises derfor kun som uddrag. Log ind for at se det fulde opslag eller gå videre til opslaget her:
