In addition to structured query operator based queries on the metadata of aerial drone images, semantic and bm25 search on the text descriptions associated with the images and vector queries of an image from the collection with the rest of the images for similarity, we can also do an object search from the detected objects for images containing those objects with the sample code as follows: 

#! /usr/bin/python 
import json 
import sys 
import os 
import requests 
import numpy as np 
from azure.core.credentials import AzureKeyCredential 
from azure.identity import DefaultAzureCredential 
from azure.search.documents import SearchClient   
from azure.search.documents.indexes import SearchIndexClient   
 
from azure.search.documents.models import ( 
    QueryCaptionResult, 
    QueryType, 
    VectorizedQuery, 
    VectorQuery, 
) 
sys.path.insert(0, os.path.abspath("..")) 
from visionprocessor.vectorizer import vectorize_image 
search_endpoint = os.getenv("AZURE_SEARCH_SERVICE_ENDPOINT")   
index_name = os.getenv("AZURE_SEARCH_INDEX_NAME") 
api_version = os.getenv("AZURE_SEARCH_API_VERSION") 
search_api_key = os.getenv("AZURE_SEARCH_ADMIN_KEY") 
vision_api_key = os.getenv("AZURE_AI_VISION_API_KEY") 
vision_api_version = os.getenv("AZURE_AI_VISION_API_VERSION") 
vision_region = os.getenv("AZURE_AI_VISION_REGION") 
vision_endpoint =  os.getenv("AZURE_AI_VISION_ENDPOINT") 
red_car_sas_url = os.getenv("AZURE_RED_CAR_SAS_URL").strip('"') 
credential = AzureKeyCredential(search_api_key) 
 
search_client = SearchClient(endpoint=search_endpoint, index_name=index_name, credential=credential) 
vector = vectorize_image(red_car_sas_url, vision_api_key, "eastus") 
vector = np.pad(vector, (0, 1536 - len(vector)), mode='constant') 
# vector_query = VectorQuery(vector=vector, 
                              # k=1000, 
                              # fields = "image_vector")   
vector_query = { 
    "vector": vector.tolist(), 
    "k": 1000,  # retrieve up to 1000 matches, adjust as needed 
    "fields": "vector", 
    "kind": "vector", 
    "exhaustive": True 
} 
results = search_client.search( 
    search_text="", 
    vector_queries= [vector_query], 
    select=["id"], 
    include_total_count=True, 
    top=1000, 
) 
if results: 
    if results.get_count() != 1000 
        print(f"Number of results: {results.get_count()}") 
    items=[] 
    for result in results: 
        if result: 
        line = result["id"] 
        id = int(line) 
        match = False 
        for item in items: 
           if item < id and id-item < 1000: 
              match = True 
              break 
           if item > id and item-id < 1000: 
              match = True 
              break 
        if match == False: 
           items+=[id] 
    for item in items: 
        print(f"{item:06d}") 
""" 
Sample output: 
008812 # invalid match no red car, all others have red cars 
011214 
015644 
006475 
010111 
007526 
017760 
012911 
014182 
016672 # only red car, full precision 
004559 
003267 
000601 
001680 
"""