app.py

import gradio as gr
import requests
import yt_dlp
import cv2
from google_img_source_search import ReverseImageSearcher
from PIL import Image
import os
import uuid

# Unique identifier for files
uid = uuid.uuid4()

# JavaScript for displaying image dimensions
size_js = """
    function imgSize(){
        var myImg = document.getElementsByClassName("my_im");
        var realWidth = myImg.naturalWidth;
        var realHeight = myImg.naturalHeight;
        alert("Original width=" + realWidth + ", " + "Original height=" + realHeight);
    }
"""

def dl(inp):
    out = None
    try:
        inp_out = inp.replace("https://", "").replace("/", "_").replace(".", "_").replace("=", "_").replace("?", "_")
        if "twitter" in inp:
            os.system(f'yt-dlp "{inp}" --extractor-arg "twitter:api=syndication" --trim-filenames 160 -o "{uid}/{inp_out}.mp4" -S res,mp4 --recode mp4')
        else:
            os.system(f'yt-dlp "{inp}" --trim-filenames 160 -o "{uid}/{inp_out}.mp4" -S res,mp4 --recode mp4')
        
        out = f"{uid}/{inp_out}.mp4"
        print(out)
    except Exception as e:
        print(e)
    return out, gr.HTML(""), "", ""

def process_vid(file, cur_frame, every_n):
    new_video_in = str(file)
    if not os.path.isfile(new_video_in):
        print(f"Video file {new_video_in} does not exist.")
        return (gr.HTML('Video file not found.'), "", "")
    
    capture = cv2.VideoCapture(new_video_in)
    frame_count = int(capture.get(cv2.CAP_PROP_FRAME_COUNT))
    rev_img_searcher = ReverseImageSearcher()
    html_out = ""
    count = int(every_n)
    start_frame = int(cur_frame) if cur_frame else 0
    
    try:
        for i in range(start_frame, frame_count - 1):
            if count == int(every_n):
                count = 1
                print(i)
                capture.set(cv2.CAP_PROP_POS_FRAMES, i)
                ret, frame_f = capture.read()
                if not ret:
                    print(f"Failed to read frame {i}")
                    continue
                
                cv2.imwrite(f"{uid}-vid_tmp{i}.png", frame_f)
                out = os.path.abspath(f"{uid}-vid_tmp{i}.png")
                out_url = f'https://nymbo-reverse-image.hf.space/file={out}'
                print(out)
                
                res = rev_img_searcher.search(out_url)
                out_cnt = 0
                if res:
                    for search_item in res:
                        out_cnt += 1
                        html_out += f"""
                        <div>
                        Title: {search_item.page_title}<br>
                        Site: <a href='{search_item.page_url}' target='_blank' rel='noopener noreferrer'>{search_item.page_url}</a><br>
                        Img: <a href='{search_item.image_url}' target='_blank' rel='noopener noreferrer'>{search_item.image_url}</a><br>
                        <img class='my_im' src='{search_item.image_url}'><br>
                        </div>"""
                    return gr.HTML(f'<h1>Total Found: {out_cnt}</h1><br>{html_out}'), f"Found frame: {i}", i + int(every_n)
            count += 1
            print(i + 1)
    except Exception as e:
        return gr.HTML(f'{e}'), "", ""
    
    return gr.HTML('No frame matches found.'), "", ""

def process_im(file, url):
    if not url.startswith("https://nymbo"):
        return url
    else:
        read_file = Image.open(file)
        read_file.save(f"{uid}-tmp.png")
        action_input = f"{uid}-tmp.png"
        out = os.path.abspath(action_input)
        out_url = f'https://nymbo-reverse-image.hf.space/file={out}'
        return out_url

def rev_im(image):
    html_out = ""
    image = cv2.imread(image)
    if image is None:
        return gr.HTML("Failed to load image.")
    
    cv2.imwrite(f"{uid}-im_tmp.png", image)
    out = os.path.abspath(f"{uid}-im_tmp.png")
    out_url = f'https://nymbo-reverse-image.hf.space/file={out}'    
    rev_img_searcher = ReverseImageSearcher()
    res = rev_img_searcher.search(out_url)
    count = 0
    for search_item in res:
        count += 1
        html_out += f"""
        <div>
        Title: {search_item.page_title}<br>
        Site: <a href='{search_item.page_url}' target='_blank' rel='noopener noreferrer'>{search_item.page_url}</a><br>
        Img: <a href='{search_item.image_url}' target='_blank' rel='noopener noreferrer'>{search_item.image_url}</a><br>
        <img class='my_im' src='{search_item.image_url}'><br>
        </div>"""
        
    return gr.HTML(f'<h1>Total Found: {count}</h1><br>{html_out}')

with gr.Blocks() as app:
    with gr.Row():
        gr.Column()
        with gr.Column():
            source_tog = gr.Radio(choices=["Image", "Video"], value="Image")
            with gr.Box(visible=True) as im_box:
                inp_url = gr.Textbox(label="Image URL")
                load_im_btn = gr.Button("Load Image")
                inp_im = gr.Image(label="Search Image", type='filepath')
                go_btn_im = gr.Button()
            with gr.Box(visible=False) as vid_box:
                vid_url = gr.Textbox(label="Video URL")
                vid_url_btn = gr.Button("Load URL")
                inp_vid = gr.Video(label="Search Video")
                with gr.Row():
                    every_n = gr.Number(label="Every /nth frame", value=10)
                stat_box = gr.Textbox(label="Status")
                with gr.Row():
                    go_btn_vid = gr.Button("Start")
                    next_btn = gr.Button("Next")
                
        gr.Column()
    with gr.Row():
        html_out = gr.HTML("")
    with gr.Row(visible=False):
        hid_box = gr.Textbox()

    def shuf(tog):
        return (gr.update(visible=True), gr.update(visible=False)) if tog == "Image" else (gr.update(visible=False), gr.update(visible=True))
    
    def load_image(url):
        return url
    
    im_load = load_im_btn.click(load_image, inp_url, inp_im)
    next_btn.click(process_vid, [inp_vid, hid_box, every_n], [html_out, stat_box, hid_box])
    vid_load = vid_url_btn.click(dl, vid_url, [inp_vid, html_out, stat_box, hid_box])
    vid_proc = go_btn_vid.click(process_vid, [inp_vid, hid_box, every_n], [html_out, stat_box, hid_box])
    im_proc = go_btn_im.click(rev_im, inp_im, [html_out])
    source_tog.change(shuf, [source_tog], [im_box, vid_box], cancels=[vid_proc, im_proc, im_load, vid_load])

app.queue(concurrency_count=20).launch()