Hafiz Abdul Rehman Saad

Academic

Career

Verbatim text

True Detective 
True Detective is a system for disguised face recognition in CCTV footage with respect to bare face, 
glasses, hats, fake beard and face masks as disguises. It is a deep learning based system which 
makes use of keras library and dcnn models. The data include 11 individuals [male, female and children] with 250 pictures each. The system consists of four modules. In module-1, it uses MTCNN 
for face detection. It divides the faces into aligned and un-aligned and makes two separate face 
databases. In module-2-3, it passes each database to a dcnn and extract the features from it. The 
extracted features are combined and then passed to a CNN. In module-4, the cnn then trains on these 
extracted 
features 
and 
give 
results on 
test 
data.  
Features include: 
- Realtime Face Recognition on the following: 
 Bare Face 
 Glasses 
 Hats 
 Fake Beard 
 Face Mask 
 
 
 
 
 
Technology Used: 
Keras, sklearn, Python, Deep Learning 
Google Colab 
Supervisor Name: 
Ms. Noor Ul Ain 
Group Members:   
Hafiz Abdul Rehman Saad (i17 - 0346)