In this software development project, you are required to attempt to recognize people from their facial geometric features. To test your software, you should create a database of facial images that can be obtained from an Internet image search, your own photo album, etc and extract the selected facial geometric features. Thereafter, for a selected image from this database apply modifications such as scaling, rotation, skew, changes in hue/saturation/brightness, etc to create a set of target images. Next, perform a suitable face recognition search on the facial image database for each of the target facial images that you have created. Face Recognition Techniques There are several approaches that you may use to accomplish this task. You are required to research on relevant face recognition systems pertaining to facial geometry to learn about different techniques available to train a classifier that can help match a target image against a database of images. You should first decide on which features of the face you want to use for classification. You must use a minimum of 10 features and it is encouraged that you use more features as shown in the figure. The same features must be extracted from each image to build a database of geometric features. You may accomplish this task manually, using for example a ruler and entering the features manually into
a file (the least preferred approach); using any program that allows you to read in an image and measure distances; or write your own code using a suitable API such the Face Recognizer API from Computer Vision (CV) Library to extract the feature set you want. This feature set is the enrollment data for the image to the image database.
Figure: Facial geometry measurements You will then create two classifiers, which take these features as input. One classifier should be probabilistic in the sense that you compute some sort of statistic from the enrollment data. The other could be a classifier that you have researched and found from the literature (for example, k-th nearest neighbor). You should then train your classifiers and perform recognition on the test set. The output should be a table where the first column is the image number (as written on the image) and the second column contains the recognition result. You may any suitable programming language to accomplish this task. Evaluate both of the classifiers developed by you using the target image set and the image database. Determine the error rate for recognition.
16 freelancers están ofertando el promedio de $96 para este trabajo
Hi, I will give you engaging, well-written and informative content with impeccable language and grammar and 0 plagiarism. Please accept my offer. Thanks.