Reverse Image Search

A simple Reverse Image Search function uses Resnet50 to extract the features of all images in the directory, and uses Milvus as a vector database to store and query similar images.

The UI is shown below:

You can run the following script to start the Gradio web UI:

python3 reverse_image_search.py

A brief description video is below:
https://www.bilibili.com/video/BV1oi421m7mR

中文解释文档在这里。

---

If you can't run it directly, you may need to do some preparation, including but not limilited to:

• Install libaray:

pip3 install gradio pymilvus torch

• Prepare dataset:

$ cd data
$ curl -L http://github.com/towhee-io/examples/releases/download/data/reverse_image_search.zip -O
$ unzip -q -o reverse_image_search.zip

• Install vector database:

$ wget http://github.com/milvus-io/milvus/releases/download/v2.2.10/milvus-standalone-docker-compose.yml -O docker-compose.yml
$ docker-compose up -d
$ docker ps