|
|
import numpy
|
|
|
import cv2
|
|
|
import time
|
|
|
from datetime import datetime
|
|
|
import numpy
|
|
|
from PIL import Image, ImageOps
|
|
|
import json
|
|
|
import traceback
|
|
|
|
|
|
import requests
|
|
|
link = url_rstp ='rtsp://admin:lahares2024@192.168.1.64:554/streaming/channels/1'
|
|
|
model_path = '/tools/models/3D_lite.tflite'
|
|
|
|
|
|
|
|
|
frames = []
|
|
|
|
|
|
|
|
|
vid = cv2.VideoCapture(link)
|
|
|
while 1:
|
|
|
|
|
|
try:
|
|
|
_, frame = vid.read()
|
|
|
except:
|
|
|
print("Error al capturar frame")
|
|
|
else:
|
|
|
|
|
|
if frame is None:
|
|
|
print("Frame capturado es None")
|
|
|
|
|
|
else:
|
|
|
|
|
|
print("recibido")
|
|
|
print(frame[-40:-20,-40:-20,0])
|
|
|
frames.append(frame)
|
|
|
|
|
|
if len(frames) == 10:
|
|
|
|
|
|
break
|
|
|
|
|
|
time.sleep(0.2)
|
|
|
|
|
|
|
|
|
|
|
|
def format_frames(frame, output_size):
|
|
|
"""
|
|
|
Redimensionar y añadir padding a una imagen obtenida de un video.
|
|
|
|
|
|
Args:
|
|
|
frame: Imagen que necesita ser redimensionada y rellenada.
|
|
|
output_size: Tamaño en píxeles de la imagen de salida.
|
|
|
|
|
|
Return:
|
|
|
Frame formateado con padding del tamaño especificado.
|
|
|
"""
|
|
|
# Convertir el frame a float32
|
|
|
frame = frame.astype(numpy.uint8)
|
|
|
print("process")
|
|
|
print(frame.min(),frame.max())
|
|
|
time.sleep(0.5)
|
|
|
# Convertir el frame de numpy array a PIL Image
|
|
|
frame_pil = Image.fromarray(frame)
|
|
|
|
|
|
# Redimensionar y añadir padding
|
|
|
frame_padded = ImageOps.pad(frame_pil, output_size, method=Image.Resampling.BILINEAR)
|
|
|
|
|
|
# Convertir de vuelta a numpy array y normalizar a [0, 1]
|
|
|
frame_final = numpy.array(frame_padded)
|
|
|
|
|
|
print("final")
|
|
|
print("min",frame_final.min(),frame_final.max())
|
|
|
|
|
|
time.sleep(0.5)
|
|
|
frame_final /=255.0
|
|
|
|
|
|
print("final/255")
|
|
|
print("min",frame_final.min(),frame_final.max())
|
|
|
|
|
|
return frame_final
|
|
|
|
|
|
|
|
|
|
|
|
output_size = (224,224)
|
|
|
for n,frame in enumerate(frames):
|
|
|
|
|
|
frames[n] = format_frames(frame,output_size)
|
|
|
|
|
|
|
|
|
|
|
|
frames = numpy.array(frames)[...,[2,1,0]]
|
|
|
|
|
|
frames = frames.reshape((1,frames.shape[0],
|
|
|
frames.shape[1],
|
|
|
frames.shape[2],
|
|
|
frames.shape[3]))
|
|
|
|
|
|
|
|
|
url_inference = "http://38.10.105.243:8501/v1/models/cnn_video:predict"
|
|
|
|
|
|
frames = json.dumps({'instances': frames.tolist()})
|
|
|
|
|
|
|
|
|
try:
|
|
|
|
|
|
resp = requests.post(url_inference,data=frames,headers={"Content-Type'": "application/json"})
|
|
|
|
|
|
except:
|
|
|
print(traceback.format_exc())
|
|
|
|
|
|
else:
|
|
|
|
|
|
print(resp)
|
|
|
print("respuesta",resp.json()['predictions'][0][0])
|
