About the project

This is a facial recognition model. This model detects faces and extract features from that detected faces and use that features to recognise them in data set

Salient feature of this model:

• This model can work on small datasets:-In general CNN model about 50-100 images are required to recognise a face whereas this model requires only 1-2 image in dataset to recognise a face. This saves lot of space on system.
• This model can gather information of unknown faces:-The object is captured automatically by cam. Image get converted into a link and this link is used for reverse image search on web browser, a technique used in web scraping for gathering information. We can get information/metadata in form of JSON.
• This model can work on multiple faces:-Facial detection in this model can generate n no. of faces which can be recognised by probalistic hashing
• This model can work on any lightning condition:-For this feature we used haar's cascade model. First it convert an RGB image into a low light binary image and constructs arrow from one bright spot to another bright spot which makes it independent of lightning condition.

Demo video when there is low lightning conditions.

low_light.mov

Demo video when there is mid lightning conditions.

mid_light.mp4

Demo video for multiple face detection.

multiple_detection.mov

Demo video for unknown faces.

unknown_meta_data.mp4