MangoIA / src /mango_massive_batch_analysis.py
juanvilla's picture
Upload folder using huggingface_hub
04b5e38 verified
Raw History Blame Contribute Delete
2.38 kB
import cv2
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from ultralytics import YOLO
import glob
import os
class MassiveMangoAnalyzer:
def __init__(self, pixel_cm2_ratio=0.00025):
print("[*] Cargando motor masivo YOLOv26...")
self.model = YOLO('yolov8n.pt')
self.k = pixel_cm2_ratio
def calculate_ngrdi(self, img):
B, G, R = cv2.split(img.astype(np.float32))
return (G - R) / (G + R + 1e-5)
def analyze_batch(self, input_path, output_csv):
image_files = glob.glob(os.path.join(input_path, "*.jpg")) + glob.glob(os.path.join(input_path, "*.png"))
print(f"[*] Procesando {len(image_files)} imágenes del set de validación...")
all_data = []
for img_path in image_files:
img = cv2.imread(img_path)
if img is None: continue
results = self.model.predict(img, conf=0.15, verbose=False)
ngrdi = self.calculate_ngrdi(img)
for r in results:
for i, box in enumerate(r.boxes):
x1, y1, x2, y2 = map(int, box.xyxy[0])
roi_ngrdi = ngrdi[y1:y2, x1:x2]
sev = (np.sum(roi_ngrdi < -0.05) / (roi_ngrdi.size + 1)) * 100
area_cm2 = roi_ngrdi.size * self.k
all_data.append({
"Archivo": os.path.basename(img_path),
"ID_Fruto": i + 1,
"Area_cm2": round(area_cm2, 2),
"Severidad_%": round(sev, 2),
"Estado": "AFECTADO" if sev > 1.5 else "SANO"
})
df = pd.DataFrame(all_data).sort_values(by="Severidad_%")
df.to_csv(output_csv, index=False)
return df
if __name__ == "__main__":
valid_images = r"C:\Users\juanv\Desktop\Analisis mango\Train\Mango-Antracnosis.v1i.yolo26\valid\images"
output = "REPORTE_MASIVO_ANTRACO_MANGO.csv"
analyzer = MassiveMangoAnalyzer()
df_results = analyzer.analyze_batch(valid_images, output)
print(f"\n[!] ANÁLISIS COMPLETADO. Datos guardados en {output}")
print("\nResumen Estadístico:")
print(df_results.groupby('Estado').agg({'Area_cm2': 'mean', 'Severidad_%': 'mean', 'ID_Fruto': 'count'}))