1 year ago

Technology Qdrant Vector Database Langchain Python

This video shows Python code for image similarity search - using Qdrant as the Vector Database and a free vision transformer model : ViT Model 'facebook/dino-vits16'
This is all free. No need for a paid AI service such as OpenAI. Unlike many tutorials and videos, this uses a custom dataset, rather than the usual ready made one. I have used some Real Estate Property images from a Kaggle collection, but you could just as easily add your own and modify the metadata accordingly. I introduce the project at the start, and then the second half is a sequence of me demonstrating it.

🟢 The process of finding similar images
🔗 https://huggingface.co/blog/image-similarity#the-process-of-finding-similar-images

- chapters -
00:00 intro - Real Estate Image Similarity Search
02:32 making a dataset
03:27 metadata
10:21 import dataset
12:22 source image
14:12 run code
16:33 qdrant actix vector database (Docker)
18:48 hugging face usage tips

Links:
---------
Vision Transformer (ViT)
🔗 https://huggingface.co/docs/transformers/v4.36.1/en/model_doc/vit#overview
🔗 https://huggingface.co/docs/datasets/v1.11.0/loading_datasets.html

Qdrant
🔗 https://qdrant.tech/
🔗 https://qdrant.tech/documentation/concepts/search/
🔗 https://colab.research.google.com/github/qdrant/examples/blob/master/qdrant_101_image_data/04_qdrant_101_cv.ipynb

Pytorch
🔗 https://pytorch.org/get-started/locally/

If you want a fast VPS server with Python installed check out :
🟢 https://webdock.io/en?maff=wdaff--170