Reverse Image Search Based on Milvus （v2.0）

This demo is an updated version of the original image search system, replacing the original image feature extraction model VGG with Resnet 50, adding the target detection function for images, and then building a reverse image search system in combination with Milvus.

The system architecture is shown below.

Environment requirements

The following tables show recommended configurations for reverse image search. These configurations haven been tested.

Component	Recommended Configuration
CPU	Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz
Memory	32GB
OS	Ubuntu 18.04
Software	Milvus 0.10.5 pic_search_webclient 0.2.0<br /

Data source

This demo uses the PASCAL VOC image set, which contains 17125 images with 20 categories: human; animals (birds, cats, cows, dogs, horses, sheep); transportation (planes, bikes,boats, buses, cars, motorcycles, trains); household (bottles, chairs, tables, pot plants, sofas, TVs)

Dataset size: ~ 2 GB.

Download location: https://pan.baidu.com/s/1MjACqsGiLo3oiTMcxodtdA extraction code: v78m

Note: You can also use other images for testing. This system supports the following formats: .jpg and .png.

How to deploy the system

GPU method

1. Run Milvus Docker

This demo uses Milvus 0.10.5. Refer to the Install Milvus for how to run Milvus docker.

2.Install the Python packages you need

cd /image_search/webserver
pip install -r requirements.txt

3.Modify configuration file

vim  /image_search_v2/webserver/src/common/config.py

Need to modify the milvus port or milvus ip parameters corresponding to the Milvus installation

Parameter	Description	Default setting
MILVUS_HOST	milvus service ip	127.0.0.1
MILVUS_PORT	milvus service port	19512
VECTOR_DIMENSION	Dimensionality of the vectors	2048
DATA_PATH	Path to save images	/data/jpegimages
DEFAULT_TABLE	The milvus default collection	milvus_183
UPLOAD_PATH	Storage path for uploaded images	/tmp/search-images
COCO_MODEL_PATH	Path of the target detection model	/yolov3_detector/data/yolov3_darknet
YOLO_CONFIG_PATH	Profile path for target detection model	/yolov3_detector/data/yolov3_darknet/yolo.yml

4.start query service

cd  /image_search/webserver/src
python app.py

If your yolo model is not downloaded automatically, you need to go to the image_search_v2/webserver/src/yolov3_detector/data/ path and run the paprepare_model.sh script

5. Run pic-search-webclient docker

$ docker run --name zilliz_search_images_demo_web -d --rm -p 8001:80 \
-e API_URL=http://${WEBSERVER_IP}:5000 \
milvusbootcamp/pic-search-webclient:0.2.0

In the previous command, WEBSERVER_IP specifies the server IP address that runs pic-search-webserver docker.

How to perform reverse image search

After deployment, enter ${WEBCLIENT_IP}:8001 in the browser to open the interface for reverse image search. WEBCLIENT_IP specifies the server IP address that runs pic-search-webclient docker.

Enter the path to the images folder, e.g. /data/images. click Load to load the images. The following screenshot shows the loading process.

Note: After clicking the load button, the system needs to wait for a while to respond. Please do not click again

The loading process may take several minutes. The following screenshot shows the interface with images loaded.

Select an image to search.

Tested with the recommended configuration, the system can complete a reverse image search in a few seconds. To load images from other directories, specify the path in the text box.