Ship Detector

A Real-time ship detector using Tensorflow Object Detection API and OpenCV.

Requirements

• Ubuntu 16.04
• Python 3.5
• TensorFlow 1.9
• OpenCV 3.2

Getting Started

Creating virtualenv

$ cd Ship-Detector
$ virtualenv env --python=python3.5
$ source env/bin/activate

Install Dependencies

$ pip install -r requirements.txt

Binding OpenCV

$ cp /usr/local/lib/python3.5/dist-packages/cv2.cpython-35m-x86_64-linux-gnu.so \
    ~/Ship-Detector/env/lib/python3.5/site-packages

Download the frozen inference graph (ssd_mobilenet_v1_ship_15000) from the Google Drive.

Run

$ python detector.py

Optional arguments (default value):

• Video to test --video=test.mp4
• Device index of the camera --camera=None
• Saved model name --model=ssd_inception_v2_ship_15000

Using GPU

Requirements (Install reference)

• Tensorflow-GPU 1.9
• CUDA 9.0
• cuDNN 7.1.4

Extras

• Creating your own dataset
• Training your own dataset
• Testing your own models