projekt_widzenie/main.py

import streamlit as st
from process_video import segment_video, classify
from io import StringIO
import cv2 as cv
import tempfile
import os
import numpy as np
from PIL import Image
import tensorflow as tf
from crop_hand_skeleton import crop_hand
from cvzone.HandTrackingModule import HandDetector

if __name__ == "__main__":
	detector = HandDetector(maxHands=1, mode=True, detectionCon=0.7, minTrackCon=0.8)
	model = tf.keras.models.load_model('model_pred/VGG16_sign_char_detection_model')
	
	st.set_page_config(
		page_title="Projekt widzenie"
	)
	st.title("Projekt rozpoznawanie liter z alfabetu znaków migowych z wideo")

	st.write('Załaduj film')
    
	upload_movie = st.file_uploader("Wybierz film", type=["mp4"])
    
	if upload_movie:
		st.write("Film się ładuje.....")
		tfile = tempfile.NamedTemporaryFile(delete=False)
		tfile.write(upload_movie.read())
		video_cap = cv.VideoCapture(tfile.name)
		font = cv.FONT_HERSHEY_SIMPLEX
		
		result, num, frames = segment_video(video_cap, fps=1.5)
		st.write(f"Załadowano {num} klatek")
		classifications = []
		for img in result:
			img_skeleton = crop_hand(img, detector)
			img2= cv.resize(img_skeleton,dsize=(224,224))
			#breakpoint()
			img_np = np.asarray(img2)
			classification = classify(img_np[:,:,::-1], model)
			classifications.append(classification)
			cv.putText(img_skeleton, 
                		classification, 
                		(20, 50), 
                		font, 2, 
                		(255, 255, 255), 
                		6, 
                		cv.LINE_4)
                		
			st.image(img_skeleton[:,:,::-1])
		i = 0
		last_letter = ''
		text = ''
		font = cv.FONT_HERSHEY_SIMPLEX
		width, height, layers = result[0].shape
		new_video_cap = cv.VideoCapture(tfile.name)
		
		out = cv.VideoWriter("output_video.mp4",cv.VideoWriter_fourcc(*'mp4v'), 30, (300, 300))
		print(f"VIDEO CAP {result[0].shape}")
		
		while True:
			ret, frame = new_video_cap.read()
			if ret == False:
				break
			image =cv.resize(frame, [300, 300])

			image = cv.cvtColor(image, cv.COLOR_RGB2BGR)
			cv.putText(image, 
                		last_letter, 
                		(50, 50), 
                		font, 2, 
                		(255, 255, 255), 
                		6, 
                		cv.LINE_4)
			cv.imwrite(f'frames/post/{i}.jpg', image)

			if i in frames:
				print(i)
				frame_index = frames.index(i)
				letter = classifications[frame_index]
				last_letter = letter
			
			img = cv.imread(f"frames/post/{i}.jpg")
			out.write(img)
			
			i += 1	
		video_cap.release()
		new_video_cap.release()
		out.release()
		
		os.system("ffmpeg -i output_video.mp4 -vcodec libx264 output_video2.mp4")
		video_file = open('output_video2.mp4', 'rb')
		
		st.video(video_file, format="video/mp4")