xudong-ma / coms30121-image-processing-and-computer-vision

Implementing and understanding the fundamentals of image processing and computer vision. Specifically, using Haar-like features and Hough spaces.

IPCV Coursework: The Dartboard Challenge

Details of the work done for this project is detailed in COMS30121_Report.pdf and below is a summary of the code structure where all functions used are in library.py

---

Subtask 1: The Viola-Jones Object Detector

Both the Jupyter Notebook and the python file contain the same code, however, the Jupyter Notebook has some minor annotations.

---

Subtask 2: Building and Testing your own Detector

Similarly, both the Jupyter Notebook and the python file contain the same code, however, the Jupyter Notebook has some minor annotations.

---

Subtask 3: Integration with Shape Detectors

The main code is in Q3.py which runs iteratively on all the images, re-producing the results. This takes upto 50 minutes. There are 2 Jupyter Notebooks, 1 is a demo which runs the code on 1 iteration, showing the intemediary steps, the other was used to produce some graphs for the report

---

Subtask 4: Improving your Detector

we have edited the library slightly for this stage and so made a new one called libraryQ4.py which Q4.py loads from. Q4 runs iteratively through all images, it's very similar to Q3, just with different parameters.

We have also included our code for running SSD in SSD.py courtesy of superdatascience online deep learning course.

---

LOADING IMAGES:

For iterative testing, place all test images in a folder and load that path directory when prompted.

For single testing, same as above, however you will need to specify the image name when prompted.