The search algorithm of images

AI & Machine Learning, Python — incorrectly specified categories?

1500 USD

Project translated automatically. Log in or register, to view the original Hello to everyone!We need a specialist for the search algorithm for similarities of images.We see this as a combination of image hashing and the search algorithm of the nearest neighbor.Here is the basic work process: 1 .Hashing of Images: Use perceptual hashing algorithms, such as Average Hash, DCT Hash or pHash, to generate a unique hash for each image.Ideally, this hash should represent the visual content of the image in a compact form.Keep these hashes in the database along with the corresponding image identifiers.2ndSearch for Nearest Neighbour: When a user downloads an image, calculate its hash using the same algorithm.- Use a fast neighbor search algorithm such as hashing with location known as LSH or k-d trees to find the closest hash-based coincidences.- Extract the corresponding images from the database.ThreeThe realization: Python libraries such as OpenCV, imagehash and scikit-learn can be useful for image processing, hashing and searching for close neighbors.Optimize database requests and indexing for effective search.Design Ready in Figma (in Investment) The idea is to create SaaS solutions for third parties in the B2B business model as, above all, a new marketing tool.However, the goal here is to develop a prototype to demonstrate the possibilities and how it can be implemented / presented, so a test database is needed.Ideally, we find a way to automatically extract images from the Internet to then work on the extract algorithm, as its effectiveness really requires a lot of similar elements.

Proposals 8 Discussions 1

Yauheni M.

0 0

Projects -
Rating -
Rating 40

Budget: 1500 USD Deadline: 90 days

Hello to you.

Write in personal messages to clarify questions.

thank you.

Viktor Andriichuk

3 0

Projects 3
Rating -
Rating 714

Budget: 1500 USD Deadline: 15 days

Your approach to the image similarity search algorithm using perceptive hashing and the search for the nearest neighbor is quite effective and is often used in practice. Algorithms such as Average Hash, DCT Hash and pHash are well suited to detect similarities on the visual level. They are resistant to small changes in the image, making them ideal for looking for images that are visually similar but not identical. But there are other methods based on the vectorization of the image and the search for the similarity of vectors. Instead of hashes, you can use fitch vectors extracted from the images using machine-sweating or deep learning algorithms. This can improve search accuracy by increasing storage size and calculation time. There is a complete understanding of how to realize it. Ready to take.