You describe a specific image retrieval task. Usually the images are first encoded into a useful datum, that can be used in efficient search system. Good results can be achieved with for example Resnet50 embeddings, stored in a ANN-indexed vector database e.g. Faiss or Scann.

Here are some examples:

- Google Landmark Retrieval with Resnet50: https://www.kaggle.com/code/tarkanatakkan/atak-img-retrieve

- ANN Scann index search in GLAMI-1M dataset https://github.com/glami/glami-1m/blob/main/train_test_similarity/use_scann_index.ipynb

Dear Titas De ,

Recently, we have focused on solving the problems of finding similar images and at the moment have created systems for finding logos, drawings, furniture, clothing, and other goods.

https://hackernoon.com/how-to-build-an-image-search-engine-to-find-similar-images

Regards,

Shafagat

you can follow these papers

https://ieeexplore.ieee.org/document/9071334

Article A New Approach to Descriptors Generation for Image Retrieval...

Article A Decade Survey of Content Based Image Retrieval Using Deep Learning

Article Saliency Inside: Learning Attentive CNNs for Content-Based I...

Thanks Vidit Kumar , Shafagat Mahmudova and Vaclav Kosar . This is useful.

We can use the concept of Content Based Image Retrieval wherein the search is performed on the features extracted from the input query image and comparison or similarity is searched through between the features of the query image and the database images.

You can visit the link : Chapter Image Retrieval Using Neural Networks for Word Image Spotting—A Review

Thanks Naiwrita Borah