rohanbaisantry / image-clustering

This is a simple image clustering algorithm which uses KMeans for clustering and performs 3 types of vectorization using vgg16, vgg19 and resnet50 using the weights from ImageNet

Image-Clustering using KMeans (A Python3 implementation)

This is a simple unsupervised image clustering algorithm which uses KMeans for clustering and Keras applications with weights pre-trained on ImageNet for vectorization of the images.

A folder named "output" will be created and the different clusters formed using the different algorithms will be present.

Change the following variables(present in the main() function) as per your convinience:

1. number_of_clusters - The number of clusters to be created by the clustering algorithm. (default is 10)
2. data_path - This is the path of the folder that contans the different images that you want to pass to the algorithm.
3. max_examples - The max number of examples to be used for the clustering (if None, all the images in the data_path folder will be used)
4. use_imagenets - choose which keras application to use. (Choose from: "Xception", "VGG16", "VGG19", "ResNet50", "InceptionV3", "InceptionResNetV2", "DenseNet", "MobileNetV2" and "False". If False, the image will be passed as is to the clustering algorithm)
5. use_pca - choose whether to use PCA for dimentionality reduction. (choose betwwen between "True" and "False". If use_imagenets=False, then use_pca will automatically be set to False as well)

Python Modules used:
-Keras
-Theano (as backend for keras)
-os
-sys
-random
-cv2 (openCV)
-numpy
-sklearn (scikit-learn for KMeans and PCA)
-shutil

To Run: "python image_clustering.py"

Pipeline:
step 1: Set the different parameters for the model. (The Variables mentioned above)
step 2: Initialize an object of the class "image_clustering" with the parameters set in the previous step.
step 3: Call the class's load_data() function.
step 4: Call the class's get_new_imagevector() function.
step 5: Call the clustering() function.

image-clustering's Issues

ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (500, 150528)

same error

Error during prediction

There is an error thrown during prediction

pred = model1.predict(self.images)
the error is
ValueError: Error when checking input: expected input_1 to have 4 dimensions, but got array with shape (500, 150528)
I think it's handling all images as one