mirror of
https://github.com/csd4ni3l/grass_touching_captcha.git
synced 2025-11-05 05:58:09 +01:00
38 lines
1.2 KiB
Python
38 lines
1.2 KiB
Python
from constants import JINA_URL, JINA_HEADERS
|
|
import requests, dotenv, os, sys, numpy as np
|
|
|
|
def get_embedding(input: list):
|
|
headers = JINA_HEADERS
|
|
|
|
if os.path.exists(".env"):
|
|
dotenv.load_dotenv(".env")
|
|
|
|
if not "JINA_TOKEN" in os.environ:
|
|
print("Jina Token not found, exiting...")
|
|
sys.exit(1)
|
|
|
|
headers["Authorization"] = f"Bearer {os.environ['JINA_TOKEN']}"
|
|
|
|
data = {
|
|
"model": "jina-embeddings-v3",
|
|
"input": input
|
|
}
|
|
|
|
response = requests.post(JINA_URL, headers=JINA_HEADERS, json=data)
|
|
|
|
return [jina_object["embedding"] for jina_object in response.json()["data"]]
|
|
|
|
def get_grass_touching_similarity(image_url):
|
|
grass_image_embedding, grass_word_embedding = get_embedding([image_url, "hand touching grass"])
|
|
|
|
grass_image_embedding = np.array(grass_image_embedding)
|
|
grass_word_embedding = np.array(grass_word_embedding)
|
|
|
|
dot_product = np.dot(grass_image_embedding, grass_word_embedding)
|
|
|
|
grass_image_mag = np.linalg.norm(grass_image_embedding)
|
|
grass_word_mag = np.linalg.norm(grass_word_embedding)
|
|
|
|
cosine_similarity = dot_product / (grass_image_mag * grass_word_mag)
|
|
|
|
return cosine_similarity |