Water Level Detection for Ship Hull

Conducted for ENPM673 Spring 2022

Sri Sai Charan V Vaishanth Ramaraj Pranav Limbekar Yash Kulkarni Jerry Pittman, Jr.

---

RAFT

This repository contains the source code and paper for RAFT:

RAFT: Recurrent All Pairs Field Transforms for Optical Flow
ECCV 2020
Zachary Teed and Jia Deng

Requirements

The code has been tested with PyTorch 1.6 and Cuda 10.1 as well as with CPU computer.

conda create --name raft
conda activate raft
conda install pytorch=1.6.0 torchvision=0.7.0 cudatoolkit=10.1 matplotlib tensorboard scipy opencv -c pytorch

Folder Structure:

📦Draft-Marker-Reading-Water-Level-Detection
 ┣ 📂Results
 ┃ ┣ 📜Output1.png
 ┃ ┣ 📜Output_screenshot_03.05.2022.png
 ┃ ┣ 📜RAFT.png
 ┃ ┣ 📜Water_Level_Detection_Result_Frame.png
 ┃ ┣ 📜gray1.png
 ┃ ┗ 📜gray_screenshot_03.05.2022.png
 ┣ 📂alt_cuda_corr
 ┃ ┣ 📜correlation.cpp
 ┃ ┣ 📜correlation_kernel.cu
 ┃ ┗ 📜setup.py
 ┣ 📂core
 ┃ ┣ 📂utils
 ┃ ┃ ┣ 📜__init__.py
 ┃ ┃ ┣ 📜augmentor.py
 ┃ ┃ ┣ 📜flow_viz.py
 ┃ ┃ ┣ 📜frame_utils.py
 ┃ ┃ ┗ 📜utils.py
 ┃ ┣ 📜__init__.py
 ┃ ┣ 📜corr.py
 ┃ ┣ 📜datasets.py
 ┃ ┣ 📜extractor.py
 ┃ ┣ 📜raft.py
 ┃ ┗ 📜update.py
 ┣ 📂input
 ┃ ┗ 📜water_level.mp4
 ┣ 📂models
 ┃ ┗ 📜raft-things.pth
 ┣ 📜ENPM673 Project 4 presentation (Group13).pdf
 ┣ 📜README.md
 ┣ 📜cnn_training.py
 ┣ 📜main.py
 ┗ 📜waterdetection.py

To Run our Code on the Vessel Movie to detect and mark water levels

If have GPU, run code as is:

python main.py

if don't have a GPU:

1. Comment out line 20
2. Uncomment line 23
3. Run Code:

python main.py

CNN Training:

python cnn_training.py

Final Output video GDrive: https://drive.google.com/drive/folders/1S9JEo39-vkEtpP1TtJBmfUNx4ZZ1hCPv?usp=sharing

Hardware Implementation using Raspberry Pi

python3 waterdetection.py

Sadly, for the the hardware implemention that due to slow processing onboard the robot with raspberry pi the number detected in the video doesn’t change. Level detection with water especially with flow takes a great deal amount of processing.