Underwater Object (Gate) Detection using Pytorch

Overview

This repository created to train a model (Pre-trained Faster R-CNN ResNet-50 FPN Model)and detect an underwater object using pytorch API with the help of Detecto.

See works with other frameworks: tensorflow, openCV.

Our team works on an Autonomous Underwater Vehicle to prepare for Singapore AUV Challange and the RoboSub'20.

Check out our website and repo.

Tested on Ubuntu 18.04

Getting Ready

1. Install Detecto

pip3 install detecto

2. Get your images for dataset.

3. Use labelimg to label images.

pip3 install labelImg

4. Run below codes to convert xml files.

from detecto.utils import xml_to_csv
xml_to_csv('path_to_xml_folder/', 'labels.csv')

Training

Local

• Train on your PC, run this script.

Server

• Train on a powerful machine, run colab notebook on Google Colab.

Testing

• Test using openCV, run this script.