My job is to perform gesture recognition. I want to do that by training a support vector machine using the features extracted by performing PCA(Principal component Analysis).
But I'm getting a little confused about the procedure.
After going through various articles, I've figured out these steps.
Take 'd' number of images(n*n) of the same gesture.
Convert each n*n image into a sigle row.
Form a matrix of order d*(n*n).
Compute the eigen values & eigen vectors.
Use top 'k' eigen vectors to form a subspace.
Project the image from original n*n dimension to 'k' dimension.
Question:
1) I have a set of 100 gestures and performing above 6 steps will give me 100 subspaces.My testing should be done on a realtime video to find which class a gesture falls in. Onto which supspace do I project each video frame to reduce the dimension for feeding it to the classifier?
Thank you in advance.
Related
I have been working to find temporal displacement between audio signals using a spectrogram. I have a short array containing data of a sound wave (pulses at specific frequencies). Now I want to plot spectrogram from that array. I have followed this steps (Spectrogram C++ library):
It would be fairly easy to put together your own spectrogram. The steps are:
window function (fairly trivial, e.g. Hanning)
FFT (FFTW would be a good choice but if licensing is an issue then go for Kiss FFT or
similar)
calculate log magnitude of frequency domain components(trivial: log(sqrt(re * re + im * im))
Now after performing these 3 steps, I am stuck at how to plot the spectrogram from this available data? Being naive in this field, I need some clear steps ahead to plot the spectrogram.
I know that a simple spectrogram has Frequency at Y-Axis, time at X-axis and magnitude as the color intensity.
But how do I get these three things to plot the spectrogram? (I want to observe and analyze data behind spectral peaks(what's the value on Y-axis and X-axis), the main purpose of plotting spectrogram).
Regards,
Khubaib
I am new to opencv and i am trying to track some moving objects(e.g. cars) in an image. I have computed the optical flow and have used it to implement kmeans and try something like background substraction , i mean seperate moving objects from stationary. Then i have also used the intensity of the video as information . The following screenshots are from the result of the flow and the k means segmentation respectively :
The results are not good but also not bad. How could i proceed from now on ? I am thinking of trying SURF feature extraction and SURF detector . Any ideas are welcome .
It seems you are using dense optical flow. I would advice trying some feature detection (surf, fast, whatever) followed by sparse optical flow tracking(from my experience it is better than feature matching for this task). Then, once you have the feature correspondences over some frames, you can use fundamental matrix, trifocal tensor, plane+parallax or some other method to detect moving objects. You can later cluster moving objects into different motion groups that represent different objects.
Also it seems that your camera is fixed. In this case you can drop the movement detection step, and consider only tracks with enough displacement, and then do the clustering into motion groups.
I am trying to classify MRI images of brain tumors into benign and malignant using C++ and OpenCV. I am planning on using bag-of-words (BoW) method after clustering SIFT descriptors using kmeans. Meaning, I will represent each image as a histogram with the whole "codebook"/dictionary for the x-axis and their occurrence count in the image for the y-axis. These histograms will then be my input for my SVM (with RBF kernel) classifier.
However, the disadvantage of using BoW is that it ignores the spatial information of the descriptors in the image. Someone suggested to use SPM instead. I read about it and came across this link giving the following steps:
Compute K visual words from the training set and map all local features to its visual word.
For each image, initialize K multi-resolution coordinate histograms to zero. Each coordinate histogram consist of L levels and each level
i has 4^i cells that evenly partition the current image.
For each local feature (let's say its visual word ID is k) in this image, pick out the k-th coordinate histogram, and then accumulate one
count to each of the L corresponding cells in this histogram,
according to the coordinate of the local feature. The L cells are
cells where the local feature falls in in L different resolutions.
Concatenate the K multi-resolution coordinate histograms to form a final "long" histogram of the image. When concatenating, the k-th
histogram is weighted by the probability of the k-th visual word.
To compute the kernel value over two images, sum up all the cells of the intersection of their "long" histograms.
Now, I have the following questions:
What is a coordinate histogram? Doesn't a histogram just show the counts for each grouping in the x-axis? How will it provide information on the coordinates of a point?
How would I compute the probability of the k-th visual word?
What will be the use of the "kernel value" that I will get? How will I use it as input to SVM? If I understand it right, is the kernel value is used in the testing phase and not in the training phase? If yes, then how will I train my SVM?
Or do you think I don't need to burden myself with the spatial info and just stick with normal BoW for my situation(benign and malignant tumors)?
Someone please help this poor little undergraduate. You'll have my forever gratefulness if you do. If you have any clarifications, please don't hesitate to ask.
Here is the link to the actual paper, http://www.csd.uwo.ca/~olga/Courses/Fall2014/CS9840/Papers/lazebnikcvpr06b.pdf
MATLAB code is provided here http://web.engr.illinois.edu/~slazebni/research/SpatialPyramid.zip
Co-ordinate histogram (mentioned in your post) is just a sub-region in the image in which you compute the histogram. These slides explain it visually, http://web.engr.illinois.edu/~slazebni/slides/ima_poster.pdf.
You have multiple histograms here, one for each different region in the image. The probability (or the number of items would depend on the sift points in that sub-region).
I think you need to define your pyramid kernel as mentioned in the slides.
A Convolutional Neural Network may be better suited for your task if you have enough training samples. You can probably have a look at Torch or Caffe.
I want to cluster some (x,y) coordinates based on distance using agglomerative hierarchical clustering as number of clusters are not known before. Is there any library that supports this task?
Iam doing in c++ using Opencv libraries.
http://opencv-python-tutroals.readthedocs.org/en/latest/py_tutorials/py_ml/py_kmeans/py_kmeans_opencv/py_kmeans_opencv.html#kmeans-opencv
This is a link for K-Means clustering in OpenCV for Python.
Shouldn't be too hard to convert this to c++ code once you understand the logic
In Gesture Recognition Toolkit (GRT) there is a simple module for hierarchical clustering. This is a "bottom up" approach as you need, where each observation starts in its own cluster, and pairs of clusters are merged as one moves up the hierarchy.
You can train the method by:
UnlabelledData: The only thing you really need to know about the UnlabelledData class is that you must set the number of input dimensions of your dataset before you try and add samples to the training dataset.
ClassificationData:
You must set the number of input dimensions of your dataset before you try and add samples to the training dataset,
You can not use the class label of 0 when you add a new sample to your dataset. This is because the class label of 0 is reserved for the special null gesture class.
MatrixDouble: MatrixDouble is the default datatype for storing M by N dimensional data, where M is the number of rows and N is the number of columns.
Furthermore you can save or load your models from/to a file and get the clusters by getClusters().
My objective is to train an SVM and get support vectors which i can plug into opencv's HOGdescriptor for object detection.
I have gathered 4000~ positives and 15000~ negatives and I train using the SVM provided by opencv. the results give me too many false positives.(up to 20 per image) I would clip out the false positives and add them into the pool of negatives to retrain. and I would end up with even more false positives at times! I have tried adjusting L2HysThreshold of my hogdescriptor upwards to 300 without significant improvement. is my pool of positives and negatives large enough?
the SVM training is also much faster than expected. I have tried with a feature vector size of 2916 and 12996, using grayscale images and color images on separate tries. SVM training has never taken longer than 20 minutes. I use auto_train. I am new to machine learning but from what i hear training with a dataset as large as mine should take at least a day no?
I believe cvSVM is not doing much learning and according to http://opencv-users.1802565.n2.nabble.com/training-a-HOG-descriptor-td6363437.html, it is not suited for this purpose. does anyone with experience with cvSVM have more input on this?
I am considering using SVMLight http://svmlight.joachims.org/ but it looks like there isn't a way to visualize the SVM hyperplane. What are my options?
I use opencv2.4.3 and have tried the following setsups for hogdescriptor
hog.winSize = cv::Size(100,100);
hog.cellSize = cv::Size(5,5);
hog.blockSize = cv::Size(10,10);
hog.blockStride = cv::Size(5,5); //12996 feature vector
hog.winSize = cv::Size(100,100);
hog.cellSize = cv::Size(10,10);
hog.blockSize = cv::Size(20,20);
hog.blockStride = cv::Size(10,10); //2916 feature vector
Your first descriptor dimension is way too large to be any useful. To form any reliable SVM hyperplane, you need at least the same number of positive and negative samples as your descriptor dimensions. This is because ideally you need separating information in every dimension of the hyperplane.
The number of positive and negative samples should be more or less the same unless you provide your SVM trainer with a bias parameter (may not be available in cvSVM).
There is no guarantee that HOG is a good descriptor for the type of problem you are trying to solve. Can you visually confirm that the object you are trying to detect has a distinct shape with similar orientation in all samples? A single type of flower for example may have a unique shape, however many types of flowers together don't have the same unique shape. A bamboo has a unique shape but may not be distinguishable from other objects easily, or may not have the same orientation in all sample images.
cvSVM is normally not the tool used to train SVMs for OpenCV HOG. Use the binary form of SVMLight (not free for commercial purposes) or libSVM (ok for commercial purposes). Calculate HOGs for all samples using your C++/OpenCV code and write it to a text file in the correct input format for SVMLight/libSVM. Use either of the programs to train a model using linear kernel with the optimal C. Find the optimal C by searching for the best accuracy while changing C in a loop. Calculate the detector vector (a N+1 dimensional vector where N is the dimension of your descriptor) by finding all the support vectors, multiplying alpha values by each corresponding support vector, and then for each dimension adding all the resulting alpha * values to find an ND vector. As the last element add -b where b is the hyperplane bias (you can find it in the model file coming out of SVMLight/libSVM training). Feed this N+1 dimensional detector to HOGDescriptor::setSVMDetector() and use HOGDescriptor::detect() or HOGDescriptor::detectMultiScale() for detection.
I have had successful results using SVMLight to learn SVM models when training from OpenCV, but haven't used cvSVM, so can't compare.
The hogDraw function from http://vision.ucsd.edu/~pdollar/toolbox/doc/index.html will visualise your descriptor.