DLMAIWCCV02 – Computer Vision Code Collection

This repository is a course workspace with small, focused computer-vision examples (OpenCV + webcam + YOLO) and one larger mini-project for metric evaluation of monocular depth.

Repository contents

1) OpenCV basics (Notebook)

• IntroductionOpenCV.ipynb
  • Image I/O, basic transforms, drawing, filtering, etc.

Run:

jupyter notebook IntroductionOpenCV.ipynb

2) Webcam capture utilities

• webCam.py – simple webcam stream (press q to quit)
• RecordImagesFromWebCam.py – capture frames to frames/ (press w to write, q to quit)

Run:

python webCam.py
python RecordImagesFromWebCam.py

3) YOLO object detection example (Ultralytics)

• yolo_example.py – run detection on a single image
• yolo26s.pt – model checkpoint, can be obtained from https://github.com/ultralytics/ultralytics

Run:

python yolo_example.py --image_path bus.jpg --model_size yolo26s.pt --confidence_threshold 0.3

4) Subproject: Metric Evaluation of Monocular Depth Estimation

Folder:

• Metric_Evaluation_Monocular_Depth_Estimation/

Goal:

• Calibrate a webcam (intrinsics + distortion)
• Detect an ArUco marker
• Estimate marker pose and reconstruct its metric plane
• Predict monocular depth (pluggable backend; default: MiDaS)
• Backproject depth values inside the marker region, fit a plane, align scale, compute metrics

Start here:

• Metric_Evaluation_Monocular_Depth_Estimation/README.md

Minimal quickstart:

cd Metric_Evaluation_Monocular_Depth_Estimation

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# (1) capture chessboards, then calibrate
python -m src.calibration.capture_chessboard
python -m src.calibration.calibrate_camera --square_size_m 0.025

# (2) capture one evaluation image
python -m src.main --capture --capture_output data/images/capture.jpg

# (3) run evaluation
python -m src.main --image data/images/capture.jpg --marker_length_m 0.048 --depth_backend midas

Installation

Root-level examples

Install dependencies for the root-level scripts/notebook:

pip install -r requirements.txt

Metric evaluation subproject

The subproject has its own dependency file:

pip install -r Metric_Evaluation_Monocular_Depth_Estimation/requirements.txt

Troubleshooting

• Webcam not accessible: check permissions / correct camera index.
• YOLO weights: ensure yolo26s.pt exists in the repo root.
• ArUco / pose estimation: see the subproject README (OpenCV wheels differ in cv2.aruco capabilities).

License

Educational use for the DLMAIWCCV02 Computer Vision course.