RoboSub Vision - 2026 Hydromeda Camera Feed

Live dual-camera streaming from the ROV with YOLO-based crab detection for the 2026 MATE ROV Competition.

Architecture

Component	Device	Description
rov/stream.py	Raspberry Pi 4B / CM5	Streams UC-684 cameras via RTP H264, serves Pi Camera stills over TCP
topside/interface.py	Topside laptop	PySide6 GUI with dual video feeds, YOLO26 crab detection, PS4 controller

---

ROV Setup (Raspberry Pi)

Prerequisites

sudo apt install gstreamer1.0-tools gstreamer1.0-plugins-good \
  gstreamer1.0-plugins-bad gstreamer1.0-plugins-ugly \
  gstreamer1.0-libav python3-dev

sudo pip3 install -r rov/requirements.txt --break-system-packages

picamera2 is pre-installed on Raspberry Pi OS - no additional install needed.

Running

sudo python3 stream.py <topside_ip>

Find your topside IP with ip addr show. Example: sudo python3 stream.py 192.168.2.1

Optional flags:

sudo python3 stream.py <topside_ip> --photo-port 5002

Second camera

When both UC-684s are connected, check their device nodes:

ls /dev/video*

Then uncomment the t1 lines in stream.py and update the index if needed.

---

Topside Setup

Python version requirement

Python 3.12 is required. PyGObject's GStreamer bindings are broken on Python 3.14 (current Arch default). Use pyenv if your system Python is 3.14+:

pyenv install 3.12.9
pyenv local 3.12.9
python -m venv .venv

Firewall

Open required ports before running - this is required even if UFW reports inactive:

sudo ufw allow 5000/udp   # UC-684 camera 1
sudo ufw allow 5001/udp   # UC-684 camera 2
sudo ufw allow 5002/tcp   # Pi Camera stills

The install script handles this automatically if UFW is active.

---

Linux - Ubuntu / Debian

sudo apt install \
  gstreamer1.0-tools gstreamer1.0-plugins-good \
  gstreamer1.0-plugins-bad gstreamer1.0-plugins-ugly \
  gstreamer1.0-libav \
  libgirepository-2.0-dev libcairo2-dev \
  pkg-config python3-dev python3-gi python3-gst-1.0

pip install -r topside/requirements.txt

Or use the install script:

bash topside/install_script.sh

Linux - Arch

sudo pacman -S gstreamer gst-plugins-good gst-plugins-bad \
  gst-plugins-ugly gst-libav python-gobject gst-python pyenv

pyenv install 3.12.9
pyenv local 3.12.9
python -m venv .venv
source .venv/bin/activate.fish   # Fish shell
# or: source .venv/bin/activate  # Bash/Zsh

pip install -r topside/requirements.txt

macOS

brew install gstreamer gst-plugins-good gst-plugins-bad \
  gst-plugins-ugly gst-libav gobject-introspection

pip install -r topside/requirements.txt

Windows

Not recommended - PyGObject cannot be pip-installed on Windows. If required:

1. Install GStreamer (MSVC 64-bit runtime + development)
2. Add C:\gstreamer\1.0\msvc_x86_64\bin to your system PATH
3. Install Python 3.12
4. pip install PySide6 opencv-python numpy pygame ultralytics

Video streaming will not work without a separate PyGObject installation.

---

Running (topside)

Activate your venv, place crab.pt in the topside/ folder, then:

source .venv/bin/activate.fish   # Fish
# or: source .venv/bin/activate  # Bash/Zsh

python interface.py

---

YOLO Crab Detection

Detects European Green Crabs for MATE 2026 Task 2.1 (Mitigate Invasive Species). Only Green Crabs receive bounding boxes and are counted - Rock and Jonah crabs are intentionally excluded per competition rules.

To retrain the model:

# Export dataset from Roboflow as YOLOv8 format, extract to crab-dataset/
yolo detect train data=crab-dataset/data.yaml model=yolo26n.pt epochs=100 imgsz=640 device=0
cp runs/detect/train/weights/best.pt topside/crab.pt

---

Controls

Control	Action
🦀 YOLO button	Toggle crab detection on/off
🌊 Preprocess button	Toggle underwater color correction (red boost + CLAHE)
PS4 Circle (○)	Trigger Pi Camera Module 3 still capture
Capture Frame button	Save current video frame to disk

Captured frames saved to topside/captured_frames/output_<timestamp>/.

---

Hardware

Component	Details
ROV computer	Raspberry Pi 4B (testing) / CM5 (competition)
Streaming cameras	Blue Robotics UC-684 × 2 (MJPEG, 1080p @ 30fps)
Photo camera	Raspberry Pi Camera Module 3 (12MP stills)
Controller	PS4 DualShock 4
Tether	Gigabit ethernet tether → router