Running:

To run the backend, run: python3 backend.py When the server is up, enter the tests_and_examples folder to run examples...

The project structure:

backend.py : The backend server src/AI_detect.py: library that analyze text/pictures/videos using AI src/extract_images.py library for handling files and directories.

The backend.py will create the folders "extracted" and "uploads", Will store images (possibly after conversion or frames), call functions from src/AI_detect.py and returns the result, after parsing.

Shell tool

shell_tool.py, allows sorting images according to prompts. Before running, it's recomended to add the following alias (sorry windows): alias aitool="python3 shell_tool.py Run: aitool <parent directory> <string to search> <dest directory> It is recommended to use "sorted_files" as a prefix for dest-directory At the moment, videos are transformed to frames.

HTTP request:

using python requests:

response = requests.post(url, data={ "classes" : possible_classes, "fast" : "true", "filenames" : found_files}, files=files_mapping, verify=False)

"classes": a list of the classes you want CLIP to classify "fast": by default false, if set to true, the analysis on videos is done every 15 frames, which is faster but may lead to data loss.

"filanames" : a list of all the filenames "files" all the files you want to transfer, a dictionary in the following format: {"filename" : file}

example:

1. def send_request(filename : str, possible_classes, url : str):\
2.    # Define the target URL\
3.    # Send the image and classes using POST request and print the result. \
4.    with open("mold.png", "rb") as pic:  \
5.        response = requests.post(url, data={\
6.                "classes" : ["A sunflower leaf with gray mold", "A healthy sunflower or sunflower leaf"],\
7.                "fast" : "false",\
8.                'filenames' : ["mold.png"]}, \
9.                files={"mold.png" : pic})\
10.       print(response.text)