The following function imports the necessary packages to get files from the directory and the similarity calculation. We then specify which folder to target based on the input class of the uploaded image:
import osfrom scipy.spatial import distance as distfrom sklearn.metrics.pairwise import cosine_similarityif y_classes == [0]: path = 'furniture_images/val/bed'elif y_classes == [1]: path = 'furniture_images/val/chair'else: path = 'furniture_images/val/sofa'
The following function loops through each image in the test directory and converts the image into an array, which is then used to predict the feature vector using the trained model:
mindist=10000maxcosine =0i=0for filename in os.listdir(path): image_train ...
