My project involves Image recognition which plays a crucial role in the field of computer vision. Popular Deep Convolutional Neural Networks (CNN) are majorly trained and tested against large-scale image datasets such as ImageNet, CIFAR10, CIFAR100, and MNIST. As part of this project, I created a new dataset which I hope that others will contribute to in the future. At macro level my project has three components: first, I created a data set which consists of approximately 5000 images of fingers counting from 1 to 5, secondly, I designed and trained a Deep Convolutional Neural Network on my dataset which classifies finger counts with a high accuracy when validated against the test set randomly extracted from the dataset. Finally, we created a live detection platform which can classify a finger count through the webcam.
To run the code in your computer download the code "finger_live_detection.m" and train using the images which are uploaded here. You can add more images to get higher accuracy and to avoid training you can also download the pre trained model "final_train.m" to test on your WEBCAM directly. Remember that in the images the background is constant so while testing you should have constant background(preferabbly white). This code will not work without good CPU or GPU.