The INSYS team is one of 3 teams participating in the Even Semester 2021/2022 Student Independent Study/Project. The INSYS team raised the theme of face detection using the Local Binary Patterns Histograms (LBPH) method. LBPH is one of the well-known methods in recognizing an object and distinguishing objects from the background. The LBPH algorithm is a combination of Local Binary Patterns (LBP) and Histograms of Oriented Gradients (HOG) descriptors. LBP is an easy but powerful way to extract and label pixels of an image. By using LBPH, we can easily represent face images with simple vectors. The progress of the implementation of this project is explained as follows:
- Preprocessing, preparing a dataset in the form of photos used by the model for learning. The photos used are 30 photos for each class which are distinguished by id. After that, the color changes to grayscale followed by cropping and resizing according to the obtained face read results. The photo data is stored in a folder with the file name according to the user id.
Figure 1. Preprocessing Process
Figure 1 shows the process of capturing faces to be used as a dataset where we enter the face id into the console and the computer will capture our 30 faces that have been cropped, resized, and converted to grayscale.
Figure 2. Image Id Inputting Process
After the computer lifts 30 facial datasets that have been cropped, resized and converted to grayscale, the next process is inputting Image IDs as shown in Figure 2. The inputting process aims so that the computer can easily recognize the captured dataset.
Figure 3. 30 Face of the dataset The
The dataset in Figure 30 is ready to be used to train the model (training model). The face detection process is carried out using an open source classifier owned by Intel which is usually used for computer vision.
- Training, training is the process of conducting training on the model using the selected recognizer, namely LBPH. Figure 4 shows the source code for the training model based on the existing dataset.
Figure 4. Training Process
Figure 5. Successful
- Recognition Training Process, taking video in real time and recognizing each face according to the user whose face has been successfully identified and calculating the percentage match with the facial image data they have.
Figure 6. Recognition Process
Displays the facial recognition program from the system that has been trained, which means the system can recognize whose face will be recognized.
Figure 7. Name Labeling Process
In this process labeling faces according to the given id (“0” for “none” and so on adjusts), and to find out who the name of the recorded dataset is.
Figure 8. Recognition Process Completed
After the Recognition process has been labeled, the program is run, the system will recognize the face according to the naming of the face id, and display the level of accuracy for each dataset that has been recorded on the computer. At this stage the INSYS Team has reached the face recognition stage of the dataset, namely the Recognition stage, but there are still shortcomings during the recognition stage and it is necessary to update the number of datasets, to get even better accuracy.