Take a small image dataset (400-500 images with at least 224X224X3 pixels or more). The dataset must have 4 to 5 classes of images. Find the final features of those images using pre-trained ResNet-101. Take one image from each class, then find the ten nearest neighbors of that image (using the final image embedding) from all the images. Do the same for AlexNet and compare the results.
Member 1: Md Shawmoon Azad (1912374042)
Member 2: M. Zanibul Haque Shanto (1921089042)
Member 3: Md. Sajjad Hossain (1922000042)
Member 4: Mohammed Rakibul Hasan (1921798042)
As mentioned in the problem statement we have taken a small dog dataset of 455 images which contains 5 classes. Then we used prestarined ResNet101 to extract the final features. Then we have used image embedding to find the 10 nearest images. we have taken help from Deeplake'Hub' for the final image embedding. Using the Hub API the final result we have go is that for 1 chihuahua image we have got 10 similar images of chihuahua as a result.