Supervision projects

1. Supervision project with data
- 1.1. Satellite used on boat detection
- 1.2. Aquaculture

1. Supervision project with data

1.1. Satellite used on boat detection

(likely, but not restricdet to) Methods: Computer vision, Time series analysis, list of papers and data: https://github.com/jasonmanesis/Satellite-Imagery-Datasets-Containing-Ships

1.1.1. Data Capella space:

https://www.capellaspace.com/gallery/?_industry=maritime

1.1.2. Data Airbus

https://www.intelligence-airbusds.com/imagery/constellation/ used in “B. Smith, S. Chester, and Y. Coady, ‘Ship Detection in Satellite Optical Imagery’, in 2020 3rd Artificial Intelligence and Cloud Computing Conference, Kyoto Japan, Dec. 2020, pp. 11–18. doi: 10.1145/3442536.3442539.”

1.1.3. References

2019 IEEE 454 Semantic Segmentation for Ships Detection from Satellite Imager
Small Ship Detection on Optical Satellite Imagery with YOLO and YOLT https://link.springer.com/chapter/10.1007/978-3-030-39442-4_49
Lessons Learned from Kaggle’s Airbus Challenge. https://towardsdatascience.com/lessons-learned-from-kaggles-airbus-challenge-252e25c5efac
New Approaches and Tools for Ship Detection in Optical

Satellite Imagery: Peruvian Ship Dataset (PSDS) and Mini Ship Dataset (MSDS), have been generated from optical satellite images obtained from different sources. PSDS is created from 22 satellite images of PERUSAT-1 with 0.7m spatial resolution, giving a total of 1310 images. MSDS has been generated using Google Earth satellite images, generating 2993 images of 900x900 pixels. Ships are found both at sea or inshore. Finally, results of the tests using Deep Learning Algorithms such as YOLT and YOLOv4 are presented, following the approach and the proposed tools
Distribution and Speed of Recreational Boats in Danish Waters – Where Does Boat Noise Affect Wildlife?

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4185998
«Estimating recreational fishing fleet using satellite data in the Aegean and Ionian Seas”

https://www.sciencedirect.com/science/article/pii/S0165783618301917?casa_token=ecl9njtiIFQAAAAA:0i-TNyf9HJ1xNW48BbWa82xRRHMr1hzq6wXptRKFTDOBLuerm71PAsLRTocs8-IOkMb-5Oe7Bw
“The performance of a spatio-spectral template on Ikonos imagery to automatically detect small recreational boats.”

https://www.ingentaconnect.com/contentone/asprs/pers/2007/00000073/00000001/art00006?crawler=true

1.2. Aquaculture

1.2.1. Fishnet/Salmon reidentification

(likely, but not restricdet to) Methods: Siamese neural network (SNN), Extended iamese neural network (ESNN), Computer Vision

AQ1: I have tagged dataset of ~120 different fish individuals from a video clip that is some minutes long. (this has the potential to be expanded)

additional data:
- AQ2: I also have a smaller dataset with pairs of images of same salmon individual at start size and “slaugther ready” size
- AQ3: Additional dataset from aquabyte with different quality of images (similar to dataset AQ2)

1.2.2. Aquaculture structure load

(likely, but not restricdet to) Methods: Time series analysis

In this problem we have two main datasets:

Bouy dataset with enviromental data: wind, waves etc over time
Stucture load time series capturing how hard structures are being pulled from anchors.

Suporting dataset:

Dataset describing exposure level of this site in 360 degrees to calculate the effect of different wind directions.

One possible usage of this would be:

Combining this with weather observations from frost.met.no or weather forcast one can model stucture load without the need for a bouy.

1.2.3. Aquaculture operational planning

(likely, but not restricdet to) Methods: Siamese neural network (SNN), Extended iamese neural network (ESNN), CBR, Time series

from mathisen2021using This data describes operational limits/planning for two different aquaculture sites, in total aroun 1000 operations. This can be combined with weather observations and exposure levels as seen in the paper to predict operational success.