UniTn Machine Learning project 2022
The main objective of this project is to create an image search engine where a query image is fed to a model that will return the most N similar images from a gallery. The following example shows the problem of retrieving the same place based on a visual search algorithm.
a. Given the input query image, the algorithm has to be capable of matching the input query image with another gallery image depicting the same animal.
b. The expected algorithm’s output is therefore a list of ranked matches between the query image and the gallery images. An example output is depicted below where the top-k matches are reported. In this case, the algorithm correctly matched top-1 and top-2 images while the others (the ones reported) are false matches.
Implementations:
- CNN from Scratch
- All Keras Pre-trained Models
- Transfer Learning Models
- Image Search Engine
Dataset sources:
- https://www.kaggle.com/datasets/iamsouravbanerjee/animal-image-dataset-90-different-animals/datasets
- https://www.image-net.org/challenges/LSVRC/2014/index#data
- https://www.mediasearch.dev/image
Detailed Report: