[GSoC] 223 Adding Face-Verification using OpenVino Model

[ABD-01]

Greetings,

This pull request is in correspondence to the GSoC'22 Prerequisite task.
I have added a notebook implementing a method to evaluate a face verification model using face-reidentification-retail-0095 model from Open Model Zoo.

Highlights:

• Used LFWPairs class for loading data which I contributed to Torchvision. PR Link
• Showed the calculation of Evaluations metrics: Confusion Matrix(TPR, FPR, FNR, TNR), Precision, Recall, Youden's Index
• Implemented a K-fold (k=10) cross-validation
• Used the metrics to plot ROC (Receiver Operating Characteristic) to visualize the model's performance.

Most of the code written is referring to my own face-recognition based project: Face-Unlock

Remaining Work:

• Add a binder link to the repository
• Get a review and complete follow up changes
• [Optional] Any more addition to the notebook like using live feed camera and personal data to verify a person.

About Me:
My name is Muhammed Abdullah and I am an undergraduate from VNIT, Nagpur, India. I am currently studying Object Detection and Semantic Segmentation and doing an implementation of YOLO from scratch. I found the Project "Object Detection demo application to run AI on ARM-based Android devices" very intriguing and wish to collaborate. I do have quite good experience with python, PyTorch, and open-source contributions, as well as I, have some experience with Java-based Android projects.(eg: this, this)

Here is my resume to get an idea of my domains and achievements. Resume

[adrianboguszewski]

Hi @ABD-01. Thank you for your contribution! It's a very interesting notebook but unfortunately, we cannot merge it as it relates to face identification, which is against our Ethics AI policy. However, we accept it as a pre-requisite task and it doesn't change your chances in GSoC.