You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

156 lines
5.6 KiB

#!/usr/bin/env python3
import os
import redis
import requests
import boto3
import pickle
import cv2
import time
import subprocess
import json
import sys
def main():
images_dir = "edge-detected-images"
is_images_dir = os.path.isdir(images_dir)
if(is_images_dir == False):
os.mkdir(images_dir)
r = redis.Redis(host="127.0.0.1", port=6379, db=2)
activation_id = os.environ.get('__OW_ACTIVATION_ID')
params = json.loads(sys.argv[1])
# edge_detected_images = {}
edge_detected_result = []
try:
decode_activation_id = params["activation_id"]
parts = params["parts"]
for i in range(0,parts):
if os.path.exists(images_dir+'/edge_detected_image_'+str(i)+'.jpg'):
os.remove(images_dir+'/edge_detected_image_'+str(i)+'.jpg')
for i in range(0,parts):
decode_output = "decode-output-image"+decode_activation_id+"-"+str(i)
load_image = pickle.loads(r.get(decode_output))
image_name = 'Image'+str(i)+'.jpg'
with open(image_name, 'wb') as f:
f.write(load_image)
img = cv2.imread(image_name)
# height, width = img.shape[:2]
# size = os.stat(img_name).st_size
# decoded_images_sizes[img_name] = size
image= cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
canny_output = cv2.Canny(image, 80, 150)
output_image = images_dir+'/edge_detected_image_'+str(i)+'.jpg'
cv2.imwrite(output_image, canny_output)
edge_detected_result.append('edge_detected_image_'+str(i)+'.jpg')
except Exception as e: #If not running as a part of DAG workflow and implemented as a single standalone function
image_url_list = params["image_url_links"]
parts = len(image_url_list)
for i in range(0,parts):
if os.path.exists(images_dir+'/edge_detected_image_'+str(i)+'.jpg'):
os.remove(images_dir+'/edge_detected_image_'+str(i)+'.jpg')
for i in range(0,parts):
response = requests.get(image_url_list[i])
image_name = 'Image'+str(i)+'.jpg'
with open(image_name, "wb") as f:
f.write(response.content)
img = cv2.imread(image_name)
image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
canny_output = cv2.Canny(image, 80, 150)
output_image = images_dir+'/edge_detected_image_'+str(i)+'.jpg'
cv2.imwrite(output_image, canny_output)
edge_detected_result.append('edge_detected_image_'+str(i)+'.jpg')
aws_access_key_id = os.getenv('AWS_ACCESS_KEY_ID')
aws_secret_access_key = os.getenv('AWS_SECRET_ACCESS_KEY')
aws_region = os.getenv('AWS_REGION')
s3 = boto3.client('s3', aws_access_key_id=aws_access_key_id,aws_secret_access_key=aws_secret_access_key,region_name=aws_region)
bucket_name = 'dagit-store'
folder_path = images_dir
folder_name = images_dir
for subdir, dirs, files in os.walk(folder_path):
for file in files:
file_path = os.path.join(subdir, file)
s3.upload_file(file_path, bucket_name, f'{folder_name}/{file_path.split("/")[-1]}')
s3.put_object_acl(Bucket=bucket_name, Key=f'{folder_name}/{file_path.split("/")[-1]}', ACL='public-read')
url_list=[]
for image in edge_detected_result:
url = "https://dagit-store.s3.ap-south-1.amazonaws.com/"+images_dir+"/"+image
url_list.append(url)
print(json.dumps({"edge_detected_image_url_links":url_list,
"activation_id": str(activation_id),
"number_of_images": parts
}))
return({"edge_detected_image_url_links":url_list,
"activation_id": str(activation_id),
"number_of_images": parts
})
if __name__ == "__main__":
main()
# decode_activation_id = params["activation_id"]
# decoded_images_sizes = {}
# edge_detected_images = {}
# parts = params["parts"]
# for i in range(0,parts):
# img_name = images_dir+'/Image' + str(i) + '.jpg'
# img = cv2.imread(img_name)
# # height, width = img.shape[:2]
# size = os.stat(img_name).st_size
# decoded_images_sizes[img_name] = size
# image= cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# canny_output = cv2.Canny(image, 80, 150)
# filename = 'detected-edges-' + str(i) +'.jpg'
# # Saving the image
# cv2.imwrite(edge_detect__directory+"/"+filename, canny_output)
# edge_img = cv2.imread(edge_detect__directory+"/"+filename)
# # edge_height, edge_width = edge_img.shape[:2]
# edge_detected_size = os.stat(edge_detect__directory+"/"+filename).st_size
# edge_detected_images[edge_detect__directory+"/"+filename] = edge_detected_size
# current_path = os.getcwd()
# sftp.put_d(current_path+"/"+edge_detect__directory,preserve_mtime=True,remotepath=remote_upload_path)
# detected_edge_images = os.listdir(current_path+"/"+edge_detect__directory)
# print(json.dumps({ "edge_detection_output": detected_edge_images,
# "edge_detect_activation_id": str(activation_id),
# "number_of_images_processed": parts,
# "edge_detection_execution_time": exec_time,
# "decode_execution_time": decode_execution_time,
# "edge_detected_images_size": edge_detected_images,
# "decoded_images_size": decoded_images_sizes
# }))