Lab09 - YOLO - Treinamento customizado

Treinamento de modelo customizado¶

Para treinar o modelo YOLOv8 basta usar apenas o comando train. Os parâmetros necessários para essa função são:

configuração de um arquivo yaml com as informações de caminho das imagens e classes
épocas
tamanho da imagem
nome do seu modelo

Neste exemplo vamos treinar um modelo capaz de detectar buracos em ruas e estradas.

Utilizaremos o dataset Potholes Detection for YOLOV4 que foi organizado no repositório abaixo.

Os datasets padrão YOLO são compatíveis com todas as versões, mesmo este dataset originalmente ter sido utilizado para a versão V4.-

Referencia: github/michelpf/fiap-ml-visao-computacional

In [ ]:

Copied!

from ultralytics import YOLO
import ultralytics

model = YOLO('yolov11n.pt')
from ultralytics import YOLO
import ultralytics

model = YOLO('yolov11n.pt')

Clonagem e Preparação do Dataset¶

O dataset utilizado contém imagens de buracos anotadas no formato YOLO, originalmente desenvolvido para YOLOv4 mas compatível com todas as versões YOLO.

Estrutura esperada:

dataset-pothole/
├── dataset/
│   ├── train/
│   │   ├── images/
│   │   └── labels/
│   └── test/
│       ├── images/
│       └── labels/

In [ ]:

Copied!

# certifique se que o repositório do dataset está clonado
# e que o diretório 'dataset-pothole' está presente

# !git clone https://github.com/michelpf/dataset-pothole
# certifique se que o repositório do dataset está clonado
# e que o diretório 'dataset-pothole' está presente

# !git clone https://github.com/michelpf/dataset-pothole

Criação do Arquivo de Configuração¶

In [ ]:

Copied!

!touch model/configs_modelo.yaml
!touch model/configs_modelo.yaml

In [12]:

Copied!





%%writefile configs_modelo.yaml
path: 'dataset-pothole/dataset'
train: 'train/'
val: 'test/'

nc: 1
names: ["pothole"]
%%writefile configs_modelo.yaml
path: 'dataset-pothole/dataset'
train: 'train/'
val: 'test/'

nc: 1
names: ["pothole"]

Overwriting configs_modelo.yaml

In [13]:

Copied!

arquivo_config  = "configs_modelo.yaml"

arquivo_config  = "configs_modelo.yaml"

O treinamento do modelo também realiza em conjunto a validação com os dados separados do treinamento.

In [15]:

Copied!





# Configurações de treinamento
EPOCHS = 50  # Aumentado para melhor convergência
IMG_SIZE = 640  # Tamanho padrão do YOLOv8
BATCH_SIZE = 16
LEARNING_RATE = 0.01

# Hiperparâmetros específicos
hyperparameters = {
    'epochs': EPOCHS,
    'imgsz': IMG_SIZE,
    'batch': BATCH_SIZE,
    'lr0': LEARNING_RATE,
    'workers': 8,
    'patience': 10,  # Early stopping
    'save_period': 10,  # Salvar checkpoint a cada 10 épocas
    'cache': True,  # Cache das imagens
    'mosaic': 1.0,  # Data augmentation
    'mixup': 0.1,
    'copy_paste': 0.1
}


resultados = model.train(
    data=arquivo_config,
    name='yolov8_pothole_custom',
    **hyperparameters
)

# treinamento simples 

# resultados = model.train(data=arquivo_config, epochs=3, imgsz=720, name='yolov8_pothole')
# Configurações de treinamento
EPOCHS = 50  # Aumentado para melhor convergência
IMG_SIZE = 640  # Tamanho padrão do YOLOv8
BATCH_SIZE = 16
LEARNING_RATE = 0.01

# Hiperparâmetros específicos
hyperparameters = {
    'epochs': EPOCHS,
    'imgsz': IMG_SIZE,
    'batch': BATCH_SIZE,
    'lr0': LEARNING_RATE,
    'workers': 8,
    'patience': 10,  # Early stopping
    'save_period': 10,  # Salvar checkpoint a cada 10 épocas
    'cache': True,  # Cache das imagens
    'mosaic': 1.0,  # Data augmentation
    'mixup': 0.1,
    'copy_paste': 0.1
}


resultados = model.train(
    data=arquivo_config,
    name='yolov8_pothole_custom',
    **hyperparameters
)

# treinamento simples 

# resultados = model.train(data=arquivo_config, epochs=3, imgsz=720, name='yolov8_pothole')

New https://pypi.org/project/ultralytics/8.3.153 available 😃 Update with 'pip install -U ultralytics'
Ultralytics 8.3.134 🚀 Python-3.9.6 torch-2.3.0 CPU (Apple M2)
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=configs_modelo.yaml, degrees=0.0, deterministic=True, device=cpu, dfl=1.5, dnn=False, dropout=0.0, dynamic=False, embed=None, epochs=3, 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=720, 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=yolov8n.pt, momentum=0.937, mosaic=1.0, multi_scale=False, name=yolov8_pothole2, 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/yolov8_pothole2, 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=1

                   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      7360  ultralytics.nn.modules.block.C2f             [32, 32, 1, True]             
  3                  -1  1     18560  ultralytics.nn.modules.conv.Conv             [32, 64, 3, 2]                
  4                  -1  2     49664  ultralytics.nn.modules.block.C2f             [64, 64, 2, True]             
  5                  -1  1     73984  ultralytics.nn.modules.conv.Conv             [64, 128, 3, 2]               
  6                  -1  2    197632  ultralytics.nn.modules.block.C2f             [128, 128, 2, True]           
  7                  -1  1    295424  ultralytics.nn.modules.conv.Conv             [128, 256, 3, 2]              
  8                  -1  1    460288  ultralytics.nn.modules.block.C2f             [256, 256, 1, True]           
  9                  -1  1    164608  ultralytics.nn.modules.block.SPPF            [256, 256, 5]                 
 10                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 11             [-1, 6]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 12                  -1  1    148224  ultralytics.nn.modules.block.C2f             [384, 128, 1]                 
 13                  -1  1         0  torch.nn.modules.upsampling.Upsample         [None, 2, 'nearest']          
 14             [-1, 4]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 15                  -1  1     37248  ultralytics.nn.modules.block.C2f             [192, 64, 1]                  
 16                  -1  1     36992  ultralytics.nn.modules.conv.Conv             [64, 64, 3, 2]                
 17            [-1, 12]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 18                  -1  1    123648  ultralytics.nn.modules.block.C2f             [192, 128, 1]                 
 19                  -1  1    147712  ultralytics.nn.modules.conv.Conv             [128, 128, 3, 2]              
 20             [-1, 9]  1         0  ultralytics.nn.modules.conv.Concat           [1]                           
 21                  -1  1    493056  ultralytics.nn.modules.block.C2f             [384, 256, 1]                 
 22        [15, 18, 21]  1    751507  ultralytics.nn.modules.head.Detect           [1, [64, 128, 256]]           
Model summary: 129 layers, 3,011,043 parameters, 3,011,027 gradients, 8.2 GFLOPs

Transferred 319/355 items from pretrained weights
Freezing layer 'model.22.dfl.conv.weight'
WARNING ⚠️ imgsz=[720] must be multiple of max stride 32, updating to [736]
train: Fast image access ✅ (ping: 0.0±0.0 ms, read: 110.6±43.8 MB/s, size: 31.9 KB)

train: Scanning /Users/arnaldoalvesvianajunior/shift-fiap/10-lab18-yolo/dataset-pothole/dataset/train... 1562 images, 0 backgrounds, 0 corrupt: 100%|██████████| 1562/1562 [00:00<00:00, 2661.57it/s]

train: New cache created: /Users/arnaldoalvesvianajunior/shift-fiap/10-lab18-yolo/dataset-pothole/dataset/train.cache
val: Fast image access ✅ (ping: 0.0±0.0 ms, read: 121.4±22.3 MB/s, size: 31.1 KB)

val: Scanning /Users/arnaldoalvesvianajunior/shift-fiap/10-lab18-yolo/dataset-pothole/dataset/test... 421 images, 0 backgrounds, 0 corrupt: 100%|██████████| 421/421 [00:00<00:00, 2937.88it/s]

val: New cache created: /Users/arnaldoalvesvianajunior/shift-fiap/10-lab18-yolo/dataset-pothole/dataset/test.cache

Plotting labels to runs/detect/yolov8_pothole2/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=0.002, momentum=0.9) with parameter groups 57 weight(decay=0.0), 64 weight(decay=0.0005), 63 bias(decay=0.0)

2025/06/11 17:29:58 INFO mlflow.tracking.fluent: Experiment with name '/Shared/Ultralytics' does not exist. Creating a new experiment.
2025/06/11 17:29:59 INFO mlflow.tracking.fluent: Autologging successfully enabled for statsmodels.
2025/06/11 17:29:59 WARNING mlflow.utils.autologging_utils: MLflow transformers autologging is known to be compatible with 4.35.2 <= transformers <= 4.51.2, but the installed version is 4.51.3. If you encounter errors during autologging, try upgrading / downgrading transformers to a compatible version, or try upgrading MLflow.
2025/06/11 17:30:02 INFO mlflow.tracking.fluent: Autologging successfully enabled for sklearn.
2025/06/11 17:30:09 INFO mlflow.tracking.fluent: Autologging successfully enabled for keras.
2025/06/11 17:30:09 INFO mlflow.tracking.fluent: Autologging successfully enabled for tensorflow.
2025/06/11 17:30:10 INFO mlflow.tracking.fluent: Autologging successfully enabled for transformers.

MLflow: logging run_id(9f8d1c2250d941ff880ff4c5d5fa221e) to runs/mlflow
MLflow: view at http://127.0.0.1:5000 with 'mlflow server --backend-store-uri runs/mlflow'
MLflow: disable with 'yolo settings mlflow=False'
Image sizes 736 train, 736 val
Using 0 dataloader workers
Logging results to runs/detect/yolov8_pothole2
Starting training for 3 epochs...

      Epoch    GPU_mem   box_loss   cls_loss   dfl_loss  Instances       Size

        1/3         0G      2.269      3.503      2.163         52        736:   5%|▌         | 5/98 [01:48<33:38, 21.70s/it]

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
Cell In[15], line 1
----> 1 resultados = model.train(data=arquivo_config, epochs=3, imgsz=720, name='yolov8_pothole')

File ~/Library/Python/3.9/lib/python/site-packages/ultralytics/engine/model.py:793, in Model.train(self, trainer, **kwargs)
    790     self.model = self.trainer.model
    792 self.trainer.hub_session = self.session  # attach optional HUB session
--> 793 self.trainer.train()
    794 # Update model and cfg after training
    795 if RANK in {-1, 0}:

File ~/Library/Python/3.9/lib/python/site-packages/ultralytics/engine/trainer.py:211, in BaseTrainer.train(self)
    208         ddp_cleanup(self, str(file))
    210 else:
--> 211     self._do_train(world_size)

File ~/Library/Python/3.9/lib/python/site-packages/ultralytics/engine/trainer.py:399, in BaseTrainer._do_train(self, world_size)
    394     self.tloss = (
    395         (self.tloss * i + self.loss_items) / (i + 1) if self.tloss is not None else self.loss_items
    396     )
    398 # Backward
--> 399 self.scaler.scale(self.loss).backward()
    401 # Optimize - https://pytorch.org/docs/master/notes/amp_examples.html
    402 if ni - last_opt_step >= self.accumulate:

File ~/Library/Python/3.9/lib/python/site-packages/torch/_tensor.py:525, in Tensor.backward(self, gradient, retain_graph, create_graph, inputs)
    515 if has_torch_function_unary(self):
    516     return handle_torch_function(
    517         Tensor.backward,
    518         (self,),
   (...)
    523         inputs=inputs,
    524     )
--> 525 torch.autograd.backward(
    526     self, gradient, retain_graph, create_graph, inputs=inputs
    527 )

File ~/Library/Python/3.9/lib/python/site-packages/torch/autograd/__init__.py:267, in backward(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)
    262     retain_graph = create_graph
    264 # The reason we repeat the same comment below is that
    265 # some Python versions print out the first line of a multi-line function
    266 # calls in the traceback and some print out the last line
--> 267 _engine_run_backward(
    268     tensors,
    269     grad_tensors_,
    270     retain_graph,
    271     create_graph,
    272     inputs,
    273     allow_unreachable=True,
    274     accumulate_grad=True,
    275 )

File ~/Library/Python/3.9/lib/python/site-packages/torch/autograd/graph.py:744, in _engine_run_backward(t_outputs, *args, **kwargs)
    742     unregister_hooks = _register_logging_hooks_on_whole_graph(t_outputs)
    743 try:
--> 744     return Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
    745         t_outputs, *args, **kwargs
    746     )  # Calls into the C++ engine to run the backward pass
    747 finally:
    748     if attach_logging_hooks:

KeyboardInterrupt:

In [ ]:

Copied!

import cv2
import matplotlib.pyplot as plt

image = cv2.imread("runs/detect/yolov8_pothole/F1_curve.png", cv2.IMREAD_COLOR)
plt.imshow(image)
import cv2
import matplotlib.pyplot as plt

image = cv2.imread("runs/detect/yolov8_pothole/F1_curve.png", cv2.IMREAD_COLOR)
plt.imshow(image)

In [ ]:

Copied!

image = cv2.imread("runs/detect/yolov8_pothole/results.png", cv2.IMREAD_COLOR)
plt.figure(figsize=(10,5))
plt.imshow(image)
image = cv2.imread("runs/detect/yolov8_pothole/results.png", cv2.IMREAD_COLOR)
plt.figure(figsize=(10,5))
plt.imshow(image)

In [ ]:

Copied!

dir_resultados = "runs/detect/yolov8_pothole5/weights/best.pt"
dir_resultados = "runs/detect/yolov8_pothole5/weights/best.pt"

In [ ]:

Copied!

image = cv2.imread("imagens/buraco2.jpeg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.imread("imagens/buraco2.jpeg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

In [ ]:

Copied!

model = YOLO(dir_resultados)
results = model.predict(source=image, conf=0.25)
model = YOLO(dir_resultados)
results = model.predict(source=image, conf=0.25)

In [ ]:

Copied!

image_result = results[0].plot()

plt.imshow(image_result)
image_result = results[0].plot()

plt.imshow(image_result)

In [ ]:

Copied!

image = cv2.imread("imagens/buraco.jpg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
image = cv2.imread("imagens/buraco.jpg")
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

In [ ]:

Copied!

results = model.predict(source=image, conf=0.25)
results = model.predict(source=image, conf=0.25)

In [ ]:

Copied!

image_result = results[0].plot()

plt.imshow(image_result)
image_result = results[0].plot()

plt.imshow(image_result)