Emanuel Silva

Thank you very much for your attention ...
It worked with MaskFPUExceptions in link:

https://github.com/pyscripter/python4delphi/wiki/MaskFPUExceptions

Hello guys!

I'm using Delphi 10.4.2 and Python4Delphi with Python 3.9.2 Win64

I have the error "Access Violation at address in module python39.dll"

the error continues even in the property of PythonEngine.DllName = python39.dll

and

PythonEngine.DllName = C:\Users\Administrator\AppData\Local\Programs\Python\Python39

a curiosity is that my code is running in VS Studio.

can anybody help me ? Thanks

my code:

unit Unit1;

interface

uses
  Classes, SysUtils,
  Windows, Messages, Graphics, Controls, Forms, Dialogs,
  StdCtrls, ComCtrls, ExtCtrls,
  PythonEngine, Vcl.PythonGUIInputOutput, SynEditHighlighter,
  SynEditCodeFolding, SynHighlighterPython, SynEditPythonBehaviour, SynEdit;

type
  TForm1 = class(TForm)
    Memo1: TMemo;
    Panel1: TPanel;
    Button1: TButton;
    Splitter1: TSplitter;
    Button2: TButton;
    Button3: TButton;
    OpenDialog1: TOpenDialog;
    SaveDialog1: TSaveDialog;
    PythonGUIInputOutput1: TPythonGUIInputOutput;
    SynEdit1: TSynEdit;
    SynEditPythonBehaviour1: TSynEditPythonBehaviour;
    SynPythonSyn1: TSynPythonSyn;
    PythonEngine1: TPythonEngine;
    procedure Button1Click(Sender: TObject);
    procedure Button2Click(Sender: TObject);
    procedure Button3Click(Sender: TObject);
  private
    { Déclarations privées }
  public
    { Déclarations publiques }
  end;

var
  Form1: TForm1;

implementation

{$R *.DFM}

procedure TForm1.Button1Click(Sender: TObject);
begin
  GetPythonEngine.ExecString(UTF8Encode(SynEdit1.Text));
end;

procedure TForm1.Button2Click(Sender: TObject);
begin
  with OpenDialog1 do
  begin
    if Execute then
      Memo1.Lines.LoadFromFile(FileName);
  end;
end;

code python:

import sysconfig
import sys
from cv2 import cv2
import pathlib
import os
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image, ImageDraw
import random
import shutil
import zipfile
from tqdm import tqdm

import torch
import torchvision
from torch.utils.data import DataLoader, Dataset
from torchvision.datasets import ImageFolder
from torchvision import transforms
import random
from time import time
from torch.utils.data import random_split

caminho_do_modelo_no_google_drive = r'C:\Users\300.pt'
model = torch.load(caminho_do_modelo_no_google_drive, map_location='cpu')

lista_violento = list(pathlib.Path(r'C:\cam2').glob('*'))
lista_nao_violento = list(pathlib.Path(r'C:\cam1').glob('*'))

print ('Quantidade de Vídeos Violentos:',len(lista_violento))
print ('Quantidade de Vídeos Não Violentos:',len(lista_nao_violento))

transformer = transforms.Compose([
                                  transforms.Resize(150),               
                                  transforms.ToTensor(),                
                                  transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])

vetor_classes = ['non-violent', 'violent']

def retirar_frame_a_frame_e_fazer_predicoes_e_plotar(path_video, y_true):

  violent_count = 0
  non_violent_count = 0
  contador = 0
  
  vetor_probs = []
  vetor_labels = []
  vetor_imagens = []
  
  cap = cv2.VideoCapture(path_video)

  while (cap.isOpened()):
    ret, frame = cap.read()

    if (ret == False):
      break
    
    img = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB) 

    img_pil = Image.fromarray(img)
    vetor_imagens.append(img_pil)

    img_tensor = transformer(img_pil)
    
    img_tensor.unsqueeze_(0)

    resultado = model.forward(img_tensor)

    probs = torch.exp(resultado)

    probs = probs.max(1)

    probabilidade = probs[0].detach().numpy()[0]
    vetor_probs.append(probabilidade)

    label_predita = probs[1].detach().numpy()[0]
    vetor_labels.append(label_predita)

    contador += 1

    if (label_predita == 1):
      violent_count += 1
    else:
      non_violent_count += 1
    
    taxa_frames_violentos = 100*violent_count/contador
    taxa_frames_nao_violentos = 100*non_violent_count/contador
  
  plt.figure(figsize=(15, 10))
  for k in range(9):
    plt.subplot(3, 3, k+1)
    n = random.randint(0, len(vetor_probs)-1)
    img = vetor_imagens[n]
    label = vetor_labels[n]
    plt.imshow(img)
    plt.xticks([])
    plt.yticks([])
    plt.title('Real: {0}\nPredito: {1}'.format(vetor_classes[y_true], vetor_classes[label]))
  
  plt.show()

  return np.array(vetor_probs), np.array(vetor_labels), taxa_frames_violentos, taxa_frames_nao_violentos

model.to('cpu')
model.eval()

video = str(random.choice(lista_nao_violento))
print ('Para o vídeo: ' + video)

vetor_probs, vetor_labels, taxa_frames_violentos, taxa_frames_nao_violentos = retirar_frame_a_frame_e_fazer_predicoes_e_plotar(video, 1)

print ('taxa frames 1: {0:.2f}'.format(taxa_frames_nao_violentos))
print ('taxa frames 2: {0:.2f}'.format(taxa_frames_violentos))

#, sys.path sys.executable
#print(sys.version)