Spaces:
Sleeping
Sleeping
import pickle | |
from minisom import MiniSom | |
import numpy as np | |
import cv2 | |
import urllib.request | |
import uuid | |
from fastapi import FastAPI, HTTPException | |
from pydantic import BaseModel | |
from typing import List | |
class InputData(BaseModel): | |
data: str # image url | |
app = FastAPI() | |
# Funci贸n para construir el modelo manualmente | |
def build_model(): | |
with open('somlucuma.pkl', 'rb') as fid: | |
somecoli = pickle.load(fid) | |
MM = np.loadtxt('matrizMM.txt', delimiter=" ") | |
return somecoli,MM | |
som,MM = build_model() # Construir el modelo al iniciar la aplicaci贸n | |
from scipy.ndimage import median_filter | |
from scipy.signal import convolve2d | |
def sobel(patron): | |
gx = np.array([[-1, 0, 1], [-2, 0, 2], [-1, 0, 1]], dtype=np.float32) | |
gy = np.array([[1, 2, 1], [0, 0, 0], [-1, -2, -1]], dtype=np.float32) | |
Gx = convolve2d(patron, gx, mode='valid') | |
Gy = convolve2d(patron, gy, mode='valid') | |
return Gx, Gy | |
def medfilt2(G, d=3): | |
return median_filter(G, size=d) | |
def orientacion(patron, w): | |
Gx, Gy = sobel(patron) | |
Gx = medfilt2(Gx) | |
Gy = medfilt2(Gy) | |
m, n = Gx.shape | |
mOrientaciones = np.zeros((m // w, n // w), dtype=np.float32) | |
for i in range(m // w): | |
for j in range(n // w): | |
Gx_patch = Gx[i*w:(i+1)*w, j*w:(j+1)*w] | |
Gy_patch = Gy[i*w:(i+1)*w, j*w:(j+1)*w] | |
YY = np.sum(2 * Gx_patch * Gy_patch) | |
XX = np.sum(Gx_patch**2 - Gy_patch**2) | |
mOrientaciones[i, j] = (0.5 * np.arctan2(YY, XX) + np.pi / 2.0) * (18.0 / np.pi) | |
return mOrientaciones | |
def redimensionar(img, h, v): | |
return cv2.resize(img, (h, v), interpolation=cv2.INTER_AREA) | |
def prediction(som, imgurl): | |
archivo = f"/tmp/test-{uuid.uuid4()}.jpg" | |
urllib.request.urlretrieve(imgurl, archivo) | |
Xtest = redimensionar(cv2.imread(archivo),256,256) | |
Xtest = np.array(Xtest) | |
Xtest = cv2.cvtColor(Xtest, cv2.COLOR_BGR2GRAY) | |
orientaciones = orientacion(Xtest, w=14) | |
Xtest = Xtest.astype('float32') / 255.0 | |
orientaciones = orientaciones.reshape(-1) | |
return som.winner(orientaciones) | |
# Ruta de predicci贸n | |
async def predict(data: InputData): | |
print(f"Data: {data}") | |
global som | |
global MM | |
try: | |
# Convertir la lista de entrada a un array de NumPy para la predicci贸n | |
imgurl = data.data | |
print(type(data.data)) | |
w = prediction(som, imgurl) | |
return {"prediction": MM[w]} | |
except Exception as e: | |
raise HTTPException(status_code=500, detail=str(e)) |