Jyo's profile - activity

try double backslashes, String face_cascade_name = "C:\\haarcascade_frontalface_alt.xml";

@albertofernandez I just applied illumination normalization like you said and it did work a lot better! It's detecting happy faces really well, but it doesn't seem to detect neutral faces at all. If I remove my threshold, it would be detected as happy. I'll increase my number of training images and see what I can do about it. Thank you so much for your help!

@albertofernandez Thank you for the comment and article reference! It's detailed and well-written with figures and examples. I have tried with both (8,1) and (8,2) neighbourhood but both seems pretty shaky. Is there any idea how I can further increase the recognition rate? I'm using OpenCV's FaceRecognizer train and predict method with LBPH. I have around 70 subjects per facial expression class (7 classes) and they have all been cropped and aligned.

I'm not well-versed in OpenCV, but I have a working gender classification program, and I am using exactly the same method as you. I still have it at around 75% accuracy, but it can all be improved afterthat. Like what you said, I believe it is returning the most matching face too, but that doesn't mean it can't be believed.

The thing about it is to have a more diverse set of face database, it'll be good to increase the number of faces you have for both genders, with different races, different emotions (or void of any). It'll be good to include a few faces that is unique, for example, faces of male that look feminine, or faces of female that looks boyish. The larger your database, the more diverse it is, I believe it'll return better results.

20 faces for each class is pretty small, many others recommend to have a few hundreds.

There are definitely many other ways to do it, or even better ways, but what you're doing is not wrong either, and there are people who reached up to over 95% of accuracy, so if you don't have a lot of time to learn about a total new algorithm or method, I think it'll be good to stick with Fisherfaces and just add on to your face database. It'll take a lot of time too, but a more diverse and large set of training data would help in any approach you're taking in gender classification I believe.

I'm trying to implement the LBPH Algorithm for Facial Expression Recognition with OpenCV. I'm reading through the documentations, and I'm curious as of what the radius parameter represents because the sentence was broken in the documentation. Any idea what the full sentence was?

(Link) Parameters:
radius – The radius used for building the Circular Local Binary Pattern. The greater the radius, the

Also, I'm trying to retrieve back the confidence level of my results too, but it's returning too huge a value which I doubt represents the confidence. I've read from some articles that it actually returns the Euclidean Distance rather than the confidence level. Is there anyway to get the confidence?

I have a working gender classification application (detect from webcam) and I'm working on emotion detection right now. I'm classifying them using different training sets but emotion detection seem to return me the same results no matter what facial expression I give. I'll eventually need to detect age groups too, so I want to make sure that what I'm doing won't make it difficult for me to enhance it further.

Am I doing them wrongly, like should I read from the same facial database instead of reading one for gender and reading one for expressions?

My codes are referenced from the OpenCV tutorial on face recognition in videos. I simply did the same steps twice for training and recognizing. If there is another approach I can take that would work with having two training datasets, do let me know!

If you think it's my codes that is causing the error (returning the same results no matter what facial expression I give), do let me know too, so I can provide the codes to find out where is wrong.

Thank you in advance :>

---------- EDIT ----------

I have changed the algorithm used for Emotion Detection. I originally used Fisherfaces, now I'm using LBPH algorithm. The results are no longer consistent, but it is very shaky, constantly changing and not exactly accurate. Is there a way to stabilize it, or is there another algorithm more suitable for emotion detection?

I'm working on Gender Classification with Fisherfaces right now, I'll need to add on to include Age Estimation and Emotion Detection. The accuracy of Fisherfaces right now is around 70-75% and I heard that adding more training images would actually help.

I intend to increase the amount of training images, but can I ask if races matter? For example, do I need to include faces of Asians to increase the accuracy of detecting an Asian's gender accurately? If so, are there any recommendations for any face databases with a wide range of unique individuals, different races, ages and emotions?

I came across this article, and because of its detailed explanation and coding, I have started to understand how Eigenfaces and Fisherfaces work.

Hope it helps other people too!

And by simply setting the labels as 0 and 1 for female and male respectively like how Face Recognition in Video does for the IDs of different people worked.

Try using:

face_cascade.load("d:\\haarcascade_frontalface_alt.xml")

I've had this problem once too, and using double backslashes worked for me.

I'm doing a school project and I am trying to do gender detection in real time and I have started off with these two tutorials provided at OpenCV.

Face recognition in video with OpenCV, Gender classification with OpenCV

I am very new to OpenCV and C++ so there are certain things that I'm unsure of. I am actually very confused with the results of the Gender Classification as of how it can help me in detecting gender. I understand that Fisherface displays the features that differs between male and female, and also the other pop-up that displayed the average face of the male and female images, but what does the Fisherface reconstruction represent?

How can I implement gender classification using these tutorials in the same way as how "Face recognition in video with OpenCV" displayed the prediction results?

You'll prolly need to link it to the opencv/build/include/ folder before they can find your opencv2 folder. I'm not sure for DevCPP, but if you were using Visual Studio you would have to set that path in your 'additional include directories' under properties. Hope something comes in your mind in regards to linking to that path for DevCPP?

I hope to identify someone's gender through a webcam, and I'm starting off with this tutorial at OpenCV: http://docs.opencv.org/trunk/modules/contrib/doc/facerec/tutorial/facerec_gender_classification.html

I simply used the demo source codes and did not do any modifications yet. After running the .exe with facerec_fisherfaces.exe C:\list.csv, the following exception occurred.

Unhandled exception at 0x6df507b9 in facerec_fisherfaces.exe: 0xC0000005: Access violation writing location 0x0c0021fe.

I am coding in C++, Visual Studio 2010 with OpenCV 2.4.7. Can I ask what and where went wrong?