** METODO Ultralytics**
In [ ]:
# Comprueba GPU
!nvidia-smi
# Paquetes base
%pip -q install ultralytics opencv-python matplotlib pandas tqdm
In [ ]:
from ultralytics import YOLO
# Load a model
model = YOLO("yolo11n.pt") # load a pretrained model (recommended for training)
# Train the model
results = model.train(data="african-wildlife.yaml", epochs=10, imgsz=640)
Ultralytics 8.3.178 🚀 Python-3.11.13 torch-2.6.0+cu124 CUDA:0 (Tesla T4, 15095MiB) engine/trainer: agnostic_nms=False, amp=True, augment=False, auto_augment=randaugment, batch=16, bgr=0.0, box=7.5, cache=False, cfg=None, classes=None, close_mosaic=10, cls=0.5, conf=None, copy_paste=0.0, copy_paste_mode=flip, cos_lr=False, cutmix=0.0, data=african-wildlife.yaml, degrees=0.0, deterministic=True, device=None, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, epochs=10, erasing=0.4, exist_ok=False, fliplr=0.5, flipud=0.0, format=torchscript, fraction=1.0, freeze=None, half=False, hsv_h=0.015, hsv_s=0.7, hsv_v=0.4, imgsz=640, int8=False, iou=0.7, keras=False, kobj=1.0, line_width=None, lr0=0.01, lrf=0.01, mask_ratio=4, max_det=300, mixup=0.0, mode=train, model=yolo11n.pt, momentum=0.937, mosaic=1.0, multi_scale=False, name=train7, nbs=64, nms=False, opset=None, optimize=False, optimizer=auto, overlap_mask=True, patience=100, perspective=0.0, plots=True, pose=12.0, pretrained=True, profile=False, project=None, rect=False, resume=False, retina_masks=False, save=True, save_conf=False, save_crop=False, save_dir=runs/detect/train7, save_frames=False, save_json=False, save_period=-1, save_txt=False, scale=0.5, seed=0, shear=0.0, show=False, show_boxes=True, show_conf=True, show_labels=True, simplify=True, single_cls=False, source=None, split=val, stream_buffer=False, task=detect, time=None, tracker=botsort.yaml, translate=0.1, val=True, verbose=True, vid_stride=1, visualize=False, warmup_bias_lr=0.1, warmup_epochs=3.0, warmup_momentum=0.8, weight_decay=0.0005, workers=8, workspace=None Overriding model.yaml nc=80 with nc=4 from n params module arguments 0 -1 1 464 ultralytics.nn.modules.conv.Conv [3, 16, 3, 2] 1 -1 1 4672 ultralytics.nn.modules.conv.Conv [16, 32, 3, 2] 2 -1 1 6640 ultralytics.nn.modules.block.C3k2 [32, 64, 1, False, 0.25] 3 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2] 4 -1 1 26080 ultralytics.nn.modules.block.C3k2 [64, 128, 1, False, 0.25] 5 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2] 6 -1 1 87040 ultralytics.nn.modules.block.C3k2 [128, 128, 1, True] 7 -1 1 295424 ultralytics.nn.modules.conv.Conv [128, 256, 3, 2] 8 -1 1 346112 ultralytics.nn.modules.block.C3k2 [256, 256, 1, True] 9 -1 1 164608 ultralytics.nn.modules.block.SPPF [256, 256, 5] 10 -1 1 249728 ultralytics.nn.modules.block.C2PSA [256, 256, 1] 11 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 12 [-1, 6] 1 0 ultralytics.nn.modules.conv.Concat [1] 13 -1 1 111296 ultralytics.nn.modules.block.C3k2 [384, 128, 1, False] 14 -1 1 0 torch.nn.modules.upsampling.Upsample [None, 2, 'nearest'] 15 [-1, 4] 1 0 ultralytics.nn.modules.conv.Concat [1] 16 -1 1 32096 ultralytics.nn.modules.block.C3k2 [256, 64, 1, False] 17 -1 1 36992 ultralytics.nn.modules.conv.Conv [64, 64, 3, 2] 18 [-1, 13] 1 0 ultralytics.nn.modules.conv.Concat [1] 19 -1 1 86720 ultralytics.nn.modules.block.C3k2 [192, 128, 1, False] 20 -1 1 147712 ultralytics.nn.modules.conv.Conv [128, 128, 3, 2] 21 [-1, 10] 1 0 ultralytics.nn.modules.conv.Concat [1] 22 -1 1 378880 ultralytics.nn.modules.block.C3k2 [384, 256, 1, True] 23 [16, 19, 22] 1 431452 ultralytics.nn.modules.head.Detect [4, [64, 128, 256]] YOLO11n summary: 181 layers, 2,590,620 parameters, 2,590,604 gradients, 6.4 GFLOPs Transferred 448/499 items from pretrained weights Freezing layer 'model.23.dfl.conv.weight' AMP: running Automatic Mixed Precision (AMP) checks... AMP: checks passed ✅ train: Fast image access ✅ (ping: 0.0±0.0 ms, read: 1181.5±453.6 MB/s, size: 54.6 KB)
train: Scanning /content/datasets/african-wildlife/labels/train.cache... 1052 images, 0 backgrounds, 0 corrupt: 100%|██████████| 1052/1052 [00:00<?, ?it/s]
albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01, method='weighted_average', num_output_channels=3), CLAHE(p=0.01, clip_limit=(1.0, 4.0), tile_grid_size=(8, 8))
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 251.1±135.3 MB/s, size: 42.2 KB)
val: Scanning /content/datasets/african-wildlife/labels/val.cache... 225 images, 0 backgrounds, 0 corrupt: 100%|██████████| 225/225 [00:00<?, ?it/s]
Plotting labels to runs/detect/train7/labels.jpg... optimizer: 'optimizer=auto' found, ignoring 'lr0=0.01' and 'momentum=0.937' and determining best 'optimizer', 'lr0' and 'momentum' automatically... optimizer: AdamW(lr=0.00125, momentum=0.9) with parameter groups 81 weight(decay=0.0), 88 weight(decay=0.0005), 87 bias(decay=0.0) Image sizes 640 train, 640 val Using 2 dataloader workers Logging results to runs/detect/train7 Starting training for 10 epochs... Closing dataloader mosaic albumentations: Blur(p=0.01, blur_limit=(3, 7)), MedianBlur(p=0.01, blur_limit=(3, 7)), ToGray(p=0.01, method='weighted_average', num_output_channels=3), CLAHE(p=0.01, clip_limit=(1.0, 4.0), tile_grid_size=(8, 8)) Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
1/10 2.48G 0.795 2.789 1.199 16 640: 100%|██████████| 66/66 [00:22<00:00, 2.92it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.20it/s]
all 225 379 0.918 0.0624 0.695 0.498
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
2/10 2.98G 0.9145 1.943 1.281 19 640: 100%|██████████| 66/66 [00:21<00:00, 3.13it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.54it/s]
all 225 379 0.616 0.33 0.454 0.241
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
3/10 2.98G 0.978 1.79 1.343 14 640: 100%|██████████| 66/66 [00:18<00:00, 3.49it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:02<00:00, 3.56it/s]
all 225 379 0.755 0.662 0.73 0.515
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
4/10 2.98G 0.9182 1.569 1.28 24 640: 100%|██████████| 66/66 [00:19<00:00, 3.42it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.42it/s]
all 225 379 0.735 0.644 0.726 0.521
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
5/10 2.98G 0.9066 1.405 1.255 18 640: 100%|██████████| 66/66 [00:21<00:00, 3.14it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.69it/s]
all 225 379 0.694 0.666 0.702 0.52
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
6/10 2.98G 0.8562 1.229 1.212 26 640: 100%|██████████| 66/66 [00:19<00:00, 3.35it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:03<00:00, 2.62it/s]
all 225 379 0.8 0.783 0.846 0.643
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
7/10 2.99G 0.7851 1.102 1.173 21 640: 100%|██████████| 66/66 [00:19<00:00, 3.32it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.86it/s]
all 225 379 0.837 0.715 0.847 0.663
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
8/10 2.99G 0.7577 1.006 1.15 20 640: 100%|██████████| 66/66 [00:21<00:00, 3.14it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.53it/s]
all 225 379 0.868 0.838 0.909 0.716
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
9/10 3G 0.7013 0.9383 1.107 17 640: 100%|██████████| 66/66 [00:20<00:00, 3.22it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:02<00:00, 3.31it/s]
all 225 379 0.903 0.847 0.921 0.745
Epoch GPU_mem box_loss cls_loss dfl_loss Instances Size
10/10 3.01G 0.6456 0.8156 1.063 18 640: 100%|██████████| 66/66 [00:19<00:00, 3.38it/s]
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:01<00:00, 4.50it/s]
all 225 379 0.9 0.905 0.934 0.761
10 epochs completed in 0.064 hours. Optimizer stripped from runs/detect/train7/weights/last.pt, 5.4MB Optimizer stripped from runs/detect/train7/weights/best.pt, 5.4MB Validating runs/detect/train7/weights/best.pt... Ultralytics 8.3.178 🚀 Python-3.11.13 torch-2.6.0+cu124 CUDA:0 (Tesla T4, 15095MiB) YOLO11n summary (fused): 100 layers, 2,582,932 parameters, 0 gradients, 6.3 GFLOPs
Class Images Instances Box(P R mAP50 mAP50-95): 100%|██████████| 8/8 [00:03<00:00, 2.20it/s]
all 225 379 0.9 0.905 0.934 0.762
buffalo 62 89 0.922 0.927 0.926 0.783
elephant 53 91 0.858 0.868 0.905 0.722
rhino 55 85 0.909 0.937 0.959 0.812
zebra 59 114 0.913 0.886 0.945 0.73
Speed: 0.3ms preprocess, 2.8ms inference, 0.0ms loss, 4.6ms postprocess per image
Results saved to runs/detect/train7
In [ ]:
# Load a model
model = YOLO("/content/runs/detect/train/weights/best.pt") # pretrained YOLO11n model
In [ ]:
# Run batched inference on a list of images
results = model("/content/datasets/african-wildlife/images/test/1 (128).jpg", save=True) # return a list of Results objects
image 1/1 /content/datasets/african-wildlife/images/test/1 (128).jpg: 416x640 1 buffalo, 51.8ms
Speed: 2.7ms preprocess, 51.8ms inference, 2.3ms postprocess per image at shape (1, 3, 416, 640)
Results saved to runs/detect/predict
In [ ]:
# Paquetes extra para el informe
%pip -q install python-docx pillow markdown
import os, glob, io, json, yaml, shutil, random, textwrap
from pathlib import Path
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from PIL import Image, ImageOps, ImageDraw, ImageFont
from docx import Document
from docx.shared import Inches, Pt
from docx.enum.text import WD_ALIGN_PARAGRAPH
from markdown import markdown
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 0.0/253.0 kB ? eta -:--:-- ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 253.0/253.0 kB 8.6 MB/s eta 0:00:00
In [ ]:
# Ajusta si tu run es otro
RUN_DIR = Path("runs/detect/train7")
assert RUN_DIR.exists(), f"No existe {RUN_DIR}. Revisa el nombre del run."
# Resultados de entrenamiento por época
csv_path = RUN_DIR / "results.csv"
df = pd.read_csv(csv_path)
last = df.iloc[-1].to_dict()
# Métricas clave finales
metrics = {
"precision": float(last.get("metrics/precision(B)", np.nan)),
"recall": float(last.get("metrics/recall(B)", np.nan)),
"mAP50": float(last.get("metrics/mAP50(B)", np.nan)),
"mAP50-95": float(last.get("metrics/mAP50-95(B)", np.nan)),
"box_loss": float(last.get("train/box_loss", np.nan)),
"cls_loss": float(last.get("train/cls_loss", np.nan)),
"dfl_loss": float(last.get("train/dfl_loss", np.nan)),
}
# Intenta leer resumen de validación por clase desde 'val.json' si existe
val_json = RUN_DIR / "val_predictions.json" # (depende de versión; a veces no se guarda)
per_class = []
if val_json.exists():
try:
import json
with open(val_json, "r") as f:
vj = json.load(f)
# Estructura puede variar. Si no hay datos por clase, se ignora.
except Exception as e:
print("No se pudo parsear val_predictions.json:", e)
# Imprime métricas resumen
metrics
Out[ ]:
{'precision': 0.90045,
'recall': 0.90455,
'mAP50': 0.93372,
'mAP50-95': 0.76123,
'box_loss': 0.64558,
'cls_loss': 0.81559,
'dfl_loss': 1.06295}
In [ ]:
# Curvas: P, R, mAP50, mAP50-95
plt.figure(); df["metrics/precision(B)"].plot(); plt.title("Precision (B)"); plt.xlabel("epoch"); plt.ylabel("precision"); plt.grid(True); plt.savefig(RUN_DIR/"curve_precision.png", dpi=150, bbox_inches="tight"); plt.close()
plt.figure(); df["metrics/recall(B)"].plot(); plt.title("Recall (B)"); plt.xlabel("epoch"); plt.ylabel("recall"); plt.grid(True); plt.savefig(RUN_DIR/"curve_recall.png", dpi=150, bbox_inches="tight"); plt.close()
plt.figure(); df["metrics/mAP50(B)"].plot(); plt.title("mAP@0.5 (B)"); plt.xlabel("epoch"); plt.ylabel("mAP50"); plt.grid(True); plt.savefig(RUN_DIR/"curve_map50.png", dpi=150, bbox_inches="tight"); plt.close()
plt.figure(); df["metrics/mAP50-95(B)"].plot(); plt.title("mAP@0.5:0.95 (B)"); plt.xlabel("epoch"); plt.ylabel("mAP50-95"); plt.grid(True); plt.savefig(RUN_DIR/"curve_map5095.png", dpi=150, bbox_inches="tight"); plt.close()
# Si Ultralytics ya generó figuras, solo las referenciamos:
pre_made_imgs = {
"labels": RUN_DIR/"labels.jpg",
"results": RUN_DIR/"results.png",
"confmat": RUN_DIR/"confusion_matrix.png", # puede no existir con algunas versiones
"pr_curve": RUN_DIR/"PR_curve.png", # idem
}
pre_made_imgs
Out[ ]:
{'labels': PosixPath('runs/detect/train7/labels.jpg'),
'results': PosixPath('runs/detect/train7/results.png'),
'confmat': PosixPath('runs/detect/train7/confusion_matrix.png'),
'pr_curve': PosixPath('runs/detect/train7/PR_curve.png')}
In [ ]:
# Usa tu mejor peso
best_pt = RUN_DIR / "weights" / "best.pt"
from ultralytics import YOLO
model = YOLO(str(best_pt))
# Busca imágenes de test
TEST_DIR = Path("/content/datasets/african-wildlife/images/test")
test_imgs = sorted(glob.glob(str(TEST_DIR/"*.jpg"))) + sorted(glob.glob(str(TEST_DIR/"*.png")))
assert len(test_imgs)>0, f"No se encontraron imágenes en {TEST_DIR}"
# Toma 12 aleatorias y predice guardando crops
sample = random.sample(test_imgs, k=min(12, len(test_imgs)))
pred_dir = RUN_DIR / "report_preds"
pred_dir.mkdir(exist_ok=True, parents=True)
# Ejecuta predicciones y guarda los PNG con anotaciones
for im_path in sample:
_ = model.predict(im_path, save=True, save_txt=False, project=str(pred_dir), name="pred", exist_ok=True)
# Ultralytics guarda en pred_dir/pred/ imagenes con las cajas
pred_imgs = sorted(glob.glob(str(pred_dir / "pred" / "*.*")))
pred_imgs[:3]
image 1/1 /content/datasets/african-wildlife/images/test/3 (209).jpg: 384x640 1 elephant, 2 rhinos, 48.7ms Speed: 2.7ms preprocess, 48.7ms inference, 1.7ms postprocess per image at shape (1, 3, 384, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/2 (366).jpg: 448x640 1 elephant, 47.3ms Speed: 2.8ms preprocess, 47.3ms inference, 1.5ms postprocess per image at shape (1, 3, 448, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/3 (54).jpg: 480x640 1 rhino, 60.0ms Speed: 2.2ms preprocess, 60.0ms inference, 1.5ms postprocess per image at shape (1, 3, 480, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/2 (122).jpg: 480x640 1 elephant, 8.6ms Speed: 1.5ms preprocess, 8.6ms inference, 1.3ms postprocess per image at shape (1, 3, 480, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/3 (240).jpg: 384x640 1 rhino, 9.0ms Speed: 1.7ms preprocess, 9.0ms inference, 1.3ms postprocess per image at shape (1, 3, 384, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/1 (356).jpg: 384x640 1 buffalo, 8.7ms Speed: 2.2ms preprocess, 8.7ms inference, 1.3ms postprocess per image at shape (1, 3, 384, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/2 (246).jpg: 384x640 2 elephants, 8.9ms Speed: 2.0ms preprocess, 8.9ms inference, 1.4ms postprocess per image at shape (1, 3, 384, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/4 (234).jpg: 576x640 1 zebra, 45.8ms Speed: 2.5ms preprocess, 45.8ms inference, 1.4ms postprocess per image at shape (1, 3, 576, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/1 (231).jpg: 448x640 1 buffalo, 10.4ms Speed: 2.2ms preprocess, 10.4ms inference, 1.3ms postprocess per image at shape (1, 3, 448, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/3 (90).jpg: 448x640 2 rhinos, 8.8ms Speed: 2.3ms preprocess, 8.8ms inference, 1.3ms postprocess per image at shape (1, 3, 448, 640) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/4 (275).jpg: 640x448 3 zebras, 45.5ms Speed: 2.0ms preprocess, 45.5ms inference, 1.5ms postprocess per image at shape (1, 3, 640, 448) Results saved to runs/detect/train7/report_preds/pred image 1/1 /content/datasets/african-wildlife/images/test/2 (182).jpg: 480x640 1 elephant, 9.2ms Speed: 2.2ms preprocess, 9.2ms inference, 1.3ms postprocess per image at shape (1, 3, 480, 640) Results saved to runs/detect/train7/report_preds/pred
Out[ ]:
['runs/detect/train7/report_preds/pred/1 (231).jpg', 'runs/detect/train7/report_preds/pred/1 (356).jpg', 'runs/detect/train7/report_preds/pred/2 (122).jpg']
In [ ]:
def make_grid(img_paths, grid=(3,4), tile_size=(640, 480), save_path=None):
cols, rows = grid[1], grid[0]
W, H = tile_size
canvas = Image.new("RGB", (cols*W, rows*H), (255,255,255))
for i, p in enumerate(img_paths[:rows*cols]):
im = Image.open(p).convert("RGB")
im = ImageOps.fit(im, (W, H))
r, c = divmod(i, cols)
canvas.paste(im, (c*W, r*H))
if save_path:
canvas.save(save_path, "PNG")
return canvas
grid_img = RUN_DIR/"grid_predictions.png"
make_grid(pred_imgs, grid=(3,4), tile_size=(640,480), save_path=grid_img)
grid_img
Out[ ]:
PosixPath('runs/detect/train7/grid_predictions.png')
In [ ]:
# Lee hparams (si existen) desde args.yaml
args_yaml = RUN_DIR / "args.yaml"
hparams = {}
if args_yaml.exists():
with open(args_yaml, "r") as f:
hparams = yaml.safe_load(f)
df_summary = pd.DataFrame({
"Métrica": ["Precision", "Recall", "mAP50", "mAP50-95", "Box loss", "Cls loss", "DFL loss"],
"Valor": [metrics["precision"], metrics["recall"], metrics["mAP50"], metrics["mAP50-95"], metrics["box_loss"], metrics["cls_loss"], metrics["dfl_loss"]]
})
df_summary
Out[ ]:
| Métrica | Valor | |
|---|---|---|
| 0 | Precision | 0.90045 |
| 1 | Recall | 0.90455 |
| 2 | mAP50 | 0.93372 |
| 3 | mAP50-95 | 0.76123 |
| 4 | Box loss | 0.64558 |
| 5 | Cls loss | 0.81559 |
| 6 | DFL loss | 1.06295 |
In [ ]:
doc = Document()
# Estilos básicos
style = doc.styles['Normal']
style.font.name = 'Calibri'
style.font.size = Pt(11)
# Portada
title = doc.add_heading('Informe de Detección de Fauna — African Wildlife (YOLO11n)', level=0)
title.alignment = WD_ALIGN_PARAGRAPH.CENTER
p = doc.add_paragraph('Experimento en Google Colab\n')
p.alignment = WD_ALIGN_PARAGRAPH.CENTER
doc.add_paragraph('Objetivos:\n'
'- Entrenar YOLO11n sobre el dataset African Wildlife (Ultralytics).\n'
'- Evaluar métricas (P, R, mAP@0.5, mAP@0.5:0.95).\n'
'- Generar visualizaciones (curvas, ejemplos de inferencia) y consolidar un informe reproducible.')
# Configuración
doc.add_heading('Configuración de entrenamiento', level=1)
if hparams:
rows = [
("epochs", hparams.get("epochs")),
("imgsz", hparams.get("imgsz")),
("batch", hparams.get("batch")),
("optimizer", hparams.get("optimizer")),
("lr0", hparams.get("lr0")),
("weight_decay", hparams.get("weight_decay")),
("device", hparams.get("device")),
("data", hparams.get("data")),
("model", hparams.get("model")),
("name", hparams.get("name")),
]
table = doc.add_table(rows=1, cols=2)
hdr = table.rows[0].cells
hdr[0].text = "Parámetro"
hdr[1].text = "Valor"
for k,v in rows:
row_cells = table.add_row().cells
row_cells[0].text = str(k)
row_cells[1].text = str(v)
else:
doc.add_paragraph("No se encontró args.yaml; se omite tabla de hiperparámetros.")
# Resultados
doc.add_heading('Métricas finales', level=1)
table = doc.add_table(rows=1, cols=2)
hdr = table.rows[0].cells
hdr[0].text = "Métrica"
hdr[1].text = "Valor"
for _, r in df_summary.iterrows():
row = table.add_row().cells
row[0].text = str(r["Métrica"])
row[1].text = f'{r["Valor"]:.4f}'
# Gráficos
doc.add_heading('Curvas y visualizaciones', level=1)
for img_path in [
RUN_DIR/"curve_precision.png",
RUN_DIR/"curve_recall.png",
RUN_DIR/"curve_map50.png",
RUN_DIR/"curve_map5095.png",
pre_made_imgs.get("labels"),
pre_made_imgs.get("results"),
pre_made_imgs.get("confmat"),
pre_made_imgs.get("pr_curve"),
RUN_DIR/"grid_predictions.png",
]:
if img_path and Path(img_path).exists():
doc.add_paragraph(Path(img_path).name)
doc.add_picture(str(img_path), width=Inches(6.0))
# Conclusiones (borrador editable)
doc.add_heading('Conclusiones (borrador)', level=1)
doc.add_paragraph(
"El modelo YOLO11n alcanzó un desempeño final de:\n"
f"- mAP@0.5 = {metrics['mAP50']:.3f}\n"
f"- mAP@0.5:0.95 = {metrics['mAP50-95']:.3f}\n"
f"con Precision={metrics['precision']:.3f} y Recall={metrics['recall']:.3f}.\n\n"
"Observaciones:\n"
"• La curva mAP muestra mejora progresiva y estabilización al final del entrenamiento.\n"
"• Las clases con menor soporte de datos pueden beneficiarse de mayor augmentación o más épocas.\n"
"• Para uso en cámaras trampa, considerar umbrales de confianza específicos por clase."
)
# Referencias
doc.add_heading('Referencias', level=1)
doc.add_paragraph(
"Ultralytics — African Wildlife Dataset: https://docs.ultralytics.com/datasets/detect/african-wildlife/\n"
"LearnOpenCV (Ankan Ghosh, 2025) — Fine-Tuning RetinaNet: https://learnopencv.com/finetuning-retinanet/\n"
"Kaggle EDA Image Datasets — Fajri (2022): https://www.kaggle.com/code/faldoae/exploratory-data-analysis-eda-for-image-datasets"
)
out_docx = RUN_DIR/"informe_yolo11n_african_wildlife.docx"
doc.save(out_docx)
out_docx
Out[ ]:
PosixPath('runs/detect/train7/informe_yolo11n_african_wildlife.docx')