This code uses Mask-RCNN (https://github.com/matterport/Mask_RCNN) as a base.
RBoxNet (rotated-bounding-box network) is a CNN to detect rotated bounding box. Its first goal is to detect objects in forward-looking sonar images.
We have tested 3 different approaches for detections of rotated bounding-box, as follows:
- Vertices
- Box offsets
- Orientation and Size
NEVES, G.; RUIZ, M.; FONTINELE, J.; OLIVEIRA, L. Rotated object detection with forward-looking sonar in underwater applications.. In: In: Elsevier Expert Systems with Applications, 2020.
Please send an e-mail to lrebouca@ufba.br to receive a link to the weights. Only valid institutional emails sent by professors linked to universities will be accepted. In the resquest email, the following text MUST be included (copy and paste the text below, filling the required fields):
"Subject: Request the weights of RBoxNet network.
- Name: [your first and last name]
- Affiliation: [university where you work]
- Department: [your department]
- Current position: [your job title]
- E-mail: [must be the e-mail at the above mentioned institution]
I have read and agreed to follow the terms and conditions below: The use of the weights of RBoxNet network is defined by the following conditions:
-
The weights are provided “AS ARE”, without any express or implied warranty. Although every effort has been made to ensure accuracy, IvisionLab does not take any responsibility for errors or omissions;
-
Without the expressed permission of IvisionLab, any of the following will be considered illegal: redistribution, modification, and commercial usage of this dataset in any way or form, either partially or in its entirety;
The weights of RBoxNet network will only be used for research purposes. I will not make any part of the networks available to a third party. I’ll not sell any part of the networks or make any profit from its use.
[your signature]"
P.S. A link to the dataset file will be sent as soon as possible.