Face Liveness Detection In Face Recognition

Face Liveness Detection is a technology in face recognition which checks whether the image from the webcam comes from a live person or not. Face Liveness Detection is an essential prerequiste of Face Recognition system.
In face recognition, the identification of an individual is done by comparing the captured images with the stored images of that person in real time. These recognition systems are in the rapid development phase and are accumulated with a new strong algorithm that improves the system day by day. However, these systems are facing many security issues as frauds are increasing on a daily basis and there is a need to upgrade these systems to make them more secure, reliable and automatic. So by observing all these things we are inspired to build our project “Face Antispoofing System” which not only improves the existing system but also adapts to some of the security challenges making this system more secure and reliable.
In this video tutorial, we perform the end to end pipeline for deploying the face antispoofing system by using opencv. First, we train the liveness detection model by using our own dataset and then we deploy the trained model using opencv and python.
Workflow of the system :
At first, the input image is captured by using a webcam and then we apply face detection algorithm to the input image. Then the detected face is forwarded to the liveness net which check whether the image comes from a real person or not.
Github link to the project files : github.com/pra...
Dataset link : drive.google.c...

[@laravelframeworkintelugu7730]

Are there any other datasets available like this. Could someone share the dataset reference if you have 😢

[@harisamin714]

the file named original vs new_dataset.png is not in dataset from where you picked this file

[@thegrasshairboi]

is this anti-spoofing system compatible with flask

[@iteducationnepal-6725]

Thank you for posting very nice content. All the best for future projects and innovations. Keep learning keep contributing for IT society🇳🇵

[@prabhatale1135]

Thank you.... You are doing great as well....

[@Raja555dy]

Last five minutes you changed,that file needed...?

[@aicogivui4646]

feasible to know, but not practical in reality. In truly face recoginition system, it's seem possible for engineer to collecting all fake images.

[@ashutoshgupta2758]

if we show video from mobile its detecting as real , how to fix that ?

[@suseeindran-qh8qy]

Training accuracy is lesser than the validation accuracy, is this fine ? I'm a newbie.

[@abderrahimbenzina5858]

waht have me original vs new_dataset.png

[@prabhatale1135]

I didnot understand what you are trying to say?

[@sunilyadav3098]

Great job. but how you collected real and fake images in the given directories.

[@KavinSRA]

Hello, what algorithm did you use?

[@hakankarapnar1859]

Hello, thank you so much for the video
can i apply RSNET-50 network instead of your model? i am newbie to deep lerning topic and i need to do something different for my project. thanks in advance

[@Damilola11199]

the liveliness code does not work on mac book

[@moihoclaptrinh2173]

hello... What algorithm does this work? and Article link ?

[@MONARKZAVERI]

man..can u share...th model that u built using ur own dataset and replacing in last five minutes?? the antispoofing model that u uploaded isnt accurate...i need th one that u build hede with lots of function...

[@Raja555dy]

Yes,I need also

[@BaoNguyen-ld2iz]

why change the hexadecimal image size to 160x160

[@zorbeyaslan8841]

I run the code. It detected my faces but it didnt work. All fake faces in photos on my tablet screen are detected as a real face!

[@prabhatale1135]

please prepare the dataset for your own and do the training as the training is done in the video... for improvement results... This is not the state of the art solution... It works for webcam and for mobile phone attack... not tested on tablet... and other use cases...

[@syedasiranjumrafi7752]

Hello, what algorithm did you use?

[@cyborg69420]

wow, great dai

[@dedisupardi2815]

Cool 👍

[@mukulmishra288]

how did you collected the spoof dataset?

[@prabhatale1135]

For live image : Use webcam to capture continuous 50 frames
For spoof image :
For photo attack: Use webcam to capture continuous 50 frames from mobile photos
For replay attack: Use a short video from mobile and start a webcam to capture these moments

[@mukulmishra288]

@@prabhatale1135is there anything we can do to automate the process of collecting data for spoof images ?

[@virugupta707]

hello brother , it is also displaying as real for photos, and in toggling between real and spoof for mobile photos.
I think the model is trained such that it detects spoof only if photos are too bright or blurry.
brother pls help me with a solution

[@MONARKZAVERI]

hey. wat was the solution for this .m facing same issue that u face....even m not able to get the th model he is building in this video...i think..th model he is buildng seems more accurate...