MATLAB Image Classification for Non-Data Scientists

It provides an image classification sample-based pre-trained deep neural network for non-data scientists. You can test the image classification by just copying images to a folder.

Requirement

It requires Deep Learning Toolbox. Pleae check Deep Learning Toolbox

It also requires to install app of pre-trained network when you use a new network.

Usage

Run demo_image_classification.

img_dir = 'images'; % specify the image folder

imds_train = load_imds( [img_dir,'/train/'] );
imds_test = load_imds( [img_dir,'/test/'] );

imcl = ImageClassifier('resnet18'); % specify the name of pre-trained netowrk.
imcl = imcl.fit( imds_train, 'num_iter', 10000, 'rho', 0.001, 'reg',1E-8, 'smooth', [0.50, 0.75] ); % parameters
[pred, proba] = imcl.pred( imds_test ); % test with test images
[results, acc] = result_table( pred, proba, imds_test ); % generate result table

Available Pre-trained feature extractor

googlenet, inceptionv3, densenet201, mobilenetv2, resnet18, resnet50, resnet101, xception, inceptionresnetv2, shufflenet, nasnetmobile, nasnetlarge, efficientnetb0, alexnet, vgg16, vgg19

Dataset

It includes four models images.

• 0: Dzee Shah That may be name of photographer. If you know the name of model, please let me know.

• 1: Robin Higgins

• 2: Saya Akane

• 3: Yuka Kawamura