im first year of computer engineering student, want to learn machine learning on facial recognition, I found deepFace sample/guide on internet, want to know if that module can save face they recognize as I can't find any mentioning about saving face id. much thanks
any references/paper or any trick to save face id
Related
I am stuck with a project on detecting suspicious facial expression along with detection of foreign objects (e.g. guns, metal rods or anything). I know nothing much of ML or image processing. I need to complete the project as soon as possible. It would be helpful if anyone could direct me with some things.
How can I manage a dataset?
Which type of code should I follow?
How do I present the final system?
I know it is a lot to ask but any amount of help is appreciated.
I have tried to train a machine using transfer learning following this link in in youtube:
https://www.youtube.com/watch?v=avv9GQ3b6Qg\
The tutorial uses mobilenet as the model and a known dataset of 7 subset (Angry, Disgust, Fear, Happy, Neutral, Sad, Surprised). I was able to successfully train the model get the face detected based on these 7 emotions.
How do I further develop it to achieve what I want?
I am trying to understand openface.
I have installed the docker container, run the demos and read the docks.
What I am missing is, how to start using it correctly.
Let me explain to you my goals:
I have an app on a raspberry pi with a webcam. If I start the app it will take a picture of the person infront.
Now it should send this picture to my openface app and check, if the face is known. Known in this context means, that I already added pictures of this person to openface before.
My questions are:
Do I need to train openface before, or could I just put the images of the persons in a directory or s.th. and compare the webcam picture on the fly with these directories?
Do I compare with images or with a generated .pkl?
Do I need to train openface for each new person?
It feels like I am missing a big thing that makes the required workflow clearer to me.
Just for the record: With help of the link I mentioned I could figure it out somehow.
Do I need to train openface before, or could I just put the images of the persons in a directory or s.th. and compare the webcam picture on the fly with these directories?
Yes, a training is required in order to compare any images.
Do I compare with images or with a generated .pkl?
Images are compared with the generated classifier pkl.
The pkl file gets generated when training openface:
./demos/classifier.py train ./generated-embeddings/
This will generate a new file called ./generated-embeddings/classifier.pkl. This file has the SVM model you'll use to recognize new faces.
Do I need to train openface for each new person?
OK, for this question I don't have an answer yet. But just because I did not look deeper into this topic yet.
I need to compare two images in a project,
The images would be two fruits of the same kind -let's say two different images of two different apples-
To be more clear, the database will have images of the stages which an apple takes from the day it was picked from a tree until it gets rotten..
The user would upload an image of the apple they have and the software should compare it to all those images in the database and retrieve the data of the matching image and tell the user at which stage is it...
I did compare before images using OpenCv emgu but I really don't have much knowledge if it's the best way...
I need an expert advise is what i said in the project even possible? or the whole database images' will match the user's image!
And is this "image processing" or something else?
And is there any suggested tutorials to learn how to do this?
I know it seems not totally clear yet, but it's just a crazy idea that I wish I can get a way to know more how i can bring it to life!
N.B the project will be an android application
This is an example of a supervised image classification problem, which is a pretty broad field. You can read up on image classification here.
The way that you would approach this problem would be to define a few stages of decay (fresh, starting to rot, half rotten, completely rotten), put together a dataset of many images of the fruit in each stage, and train an image classifier on each stage. The sample dataset should contain images of many different pieces of fruit in many different settings. If you want to support different types of fruit, you would need to train a different classifier for each fruit.
There are many image classification tools out there. To name a few:
OpenCV's haar classifier
dlib's hog classifier
Matlab's Computer Vision System Toolbox
VLFeat
It would be up to you to look into which approach would work best for your situation.
Given that this is a fairly broad problem, I wouldn't expect to come up with a solid solution quickly unless you've had experience with image classification. If you are trying to develop a product, I would recommend getting in touch with a computer vision expert that you could contract to solve it.
If you are just looking to learn more about image classification, however, this could be a fun way to play around with different tools and get a feel for what's out there. You may want to start by learning about Machine Learning in general. Caltech offers a free online course that gives a pretty good intro to the subject.
Recently, I'm studying about facial recognition with OpenCV, and I'm trying some simple example based on study.
I'm considering to use it at front door condition.
Nowadays some buildings or apartments use facial recognition for preventing intruders. When someone joins them (such as company or houses), they require the person's picture. As I know, they require just one picture.
I didn't care about that last time, but now, I'm very curious about it.
The famous algorithms such as PCA, LDA use machine learning, so they increase successful percentages(cases). To use machine learning, they need sample images as many as I can provide. That's why I'm curious about that. Buildings or companys require just one picture, but they can recognize each person. Moreover, their accuracy is very good. How can this happen? Is there any other algorithm besides PCA or LDA?
Thanks for reading!
As far as I know, this hasn't been achieved yet. So I don't think they can develop a software recognizing a person by using only one picture.
It is most likely that they teach the algorithm with the authorized person's pictures. So if that one picture does not match with the trained ones, the algorithm can say this is an intrusion.
Edit:
As linuxqwerty pointed out those commercial products are already trained with huge datasets.
As a result of this training, learning happens and the algorithm achieves feature extraction of all those sample faces.
Then the algorithm knows almost every kinds of features that an human face can have.
For example: thickness of eyebrows, distance between eyes, roundness of chin... These are only a human can say about faces. The algorithm can extract thousands of these features.
It can keep faces as a representation of those features.
So now we have this commercial software which can represent faces as binary codes with a lot of digits.
I am getting your question again.
The apartment or company bought this software.
They included the picture of authorized person.
What the software does is simply converting the picture as it was a thousand digits password.
So that person has this unique password which the system can only reproduce that password only from his face.
To sum up:
The learning part was achieved using big face databases.
Thanks to learning part, the recognition part can be done by using only one picture.
PS: Corrections are welcome.
I happened to read about facial recognition before, that time I wanted to do it as my semester project. And of course I have heard and thought of using OpenCV as well.
Your question is simple, those company or home that use facial recognition, they usually use very well-developed product, which normally includes well-programmed facial recognition. As we are talking about security here, normally companies will buy these security products, unless if they just want to use it as a tool to deter intruders which focus less on the practical usage, and recognition accuracy, they can opt for free facial recognition software.
So, when I'm talking about well-programmed facial recognition, it means that it was trained with huge amount of databases (the photos to be recognized that you mentioned), this means the training is done even before the software is officially launched, which is during the development stage. A good facial recognition software requires both good, complete and detailed programming coding, and also huge photo databases (taken at different ambient light intensity, different facial features like hair style, spectacles) to train it.
Therefore, the accuracy of the software does not depend solely on the amount of pictures given during the usage of the software provided that it is well-programmed in the first place. Thanks and hope I answered your question and wonder.
ps: recognize is spelled this way (US); recognise (UK) =)
This is a research question not a direct programming question.
I am working on a symbol recognition algorithm, What the software currently does, it takes an image, divide it into contours (blobs) and start matching each contour with a list of predefined templates. Then for each contour it takes the one that has the highest match rate.
The algorithm is doing fairely however I need to train it better. What I mean is this:
I want to use a machine learning algorithm that will train the algorithm to have better matching. So lets take an example:
I run the recognition on a symbol, the algorithm will run and find that this symbol is a car, then I have to confirm that result (maybe by clicking on "Yes" or "No") the algorithm should learn from that. So if I click on NO the algorithm should learn that this is not a car and will have better result next time (maybe try to match something else). while if i click on YES he will know that he was correct and next time he will perform better when searching for a car.
This is the concept I am trying to research. I need documents or algorithm that can achieve this sort of things. I am not looking for implementations or programming, just concept or researches.
I have done many researches and read a lot about machine learning, neural networks, decision trees.... but i was not able to know how can I use any in my scenarion.
I hope I was clear and this type of question is allowed on stack overflow. if not I'm sorry
Thanks a lot for any help or tip
Image recognition is still a challenge in the community. What you described in your process of manually clicking yes/no is just creating labeled data. Since this is a very broad area, I will just point you to a few links that might be useful.
To get start, you might want to use some existing image databases instead of creating your own, which saves you a lot of effort. e.g., this car dataset in UCIC image db.
Since you already have the background of machine learning, you can take a look at some survey paper that exactly match your project interests, e.g., search object recognition survey paper or feature extraction car in google.
Then you can dive into some good papers and see whether they are suitable for your project. For example, you can check the two papers below that linked with the UCIC image db.
Shivani Agarwal, Aatif Awan, and Dan Roth,
Learning to detect objects in images via a sparse, part-based representation.
IEEE Transactions on Pattern Analysis and Machine Intelligence, 26(11):1475-1490, 2004.
Shivani Agarwal and Dan Roth,
Learning a sparse representation for object detection.
In Proceedings of the Seventh European Conference on Computer Vision, Part IV, pages 113-130, Copenhagen, Denmark, 2002.
Also check for some implemented softwares instead of starting from scratch, in your case, opencv should be good one to start with.
For image recognition, feature extraction is one of the most important step. You might want to check some stat-of-the-art algorithms in the community. (SIFT, mean-shift, harr features etc).
Boosting algorithm might also be useful when you reach the classification step. I see a lot of scholars mention this in image recognition community.
As #nickbar suggest, discuss more at https://stats.stackexchange.com/