eniokilder/Banco-Imagem

# Projeto
Banco-Imagem

### Nome do aluno
Enio Kilder Oliveira da Silva

|**Tipo de Projeto**|**Modelo Selecionado**|**Linguagem**|
|--|--|--|
Classificação de Objetos |YOLOv5|PyTorch|

## Performance

O modelo treinado possui performance de **98.6%**.

### Output do bloco de treinamento

<details>
<summary>Expandir Conteúdo!</summary>

```text
%%time
%cd ../yolov5
!python classify/train.py --model yolov5n-cls.pt --data $DATASET_NAME --epochs 128 --batch 16 --img 320 --pretrained weights/yolov5n-cls.pt

/content/yolov5
2023-10-28 01:49:35.242300: E tensorflow/compiler/xla/stream_executor/cuda/cuda_dnn.cc:9342] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
2023-10-28 01:49:35.242363: E tensorflow/compiler/xla/stream_executor/cuda/cuda_fft.cc:609] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
2023-10-28 01:49:35.242406: E tensorflow/compiler/xla/stream_executor/cuda/cuda_blas.cc:1518] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
classify/train: model=yolov5n-cls.pt, data=Banco-Imagem-1, epochs=128, batch_size=16, imgsz=320, nosave=False, cache=None, device=, workers=8, project=runs/train-cls, name=exp, exist_ok=False, pretrained=weights/yolov5n-cls.pt, optimizer=Adam, lr0=0.001, decay=5e-05, label_smoothing=0.1, cutoff=None, dropout=None, verbose=False, seed=0, local_rank=-1
github: up to date with https://github.com/ultralytics/yolov5 ✅
YOLOv5 🚀 v7.0-230-g53efd07 Python-3.10.12 torch-2.1.0+cu118 CUDA:0 (Tesla T4, 15102MiB)

TensorBoard: Start with 'tensorboard --logdir runs/train-cls', view at http://localhost:6006/
albumentations: RandomResizedCrop(p=1.0, height=320, width=320, scale=(0.08, 1.0), ratio=(0.75, 1.3333333333333333), interpolation=1), HorizontalFlip(p=0.5), ColorJitter(p=0.5, brightness=[0.6, 1.4], contrast=[0.6, 1.4], saturation=[0.6, 1.4], hue=[0, 0]), Normalize(p=1.0, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225), max_pixel_value=255.0), ToTensorV2(always_apply=True, p=1.0, transpose_mask=False)
Downloading https://github.com/ultralytics/yolov5/releases/download/v7.0/yolov5n-cls.pt to yolov5n-cls.pt...
100% 4.87M/4.87M [00:00<00:00, 48.4MB/s]

Model summary: 149 layers, 1218405 parameters, 1218405 gradients, 3.0 GFLOPs
optimizer: Adam(lr=0.001) with parameter groups 32 weight(decay=0.0), 33 weight(decay=5e-05), 33 bias
Image sizes 320 train, 320 test
Using 1 dataloader workers
Logging results to runs/train-cls/exp
Starting yolov5n-cls.pt training on Banco-Imagem-1 dataset with 5 classes for 128 epochs...

Epoch GPU_mem train_loss test_loss top1_acc top5_acc
1/128 0.508G 1.55 1.51 0.194 1: 100% 16/16 [00:06<00:00, 2.59it/s]
2/128 0.508G 1.39 1.86 0.222 1: 100% 16/16 [00:02<00:00, 6.81it/s]
3/128 0.508G 1.4 2.07 0.194 1: 100% 16/16 [00:02<00:00, 7.04it/s]
4/128 0.508G 1.35 1.75 0.222 1: 100% 16/16 [00:02<00:00, 6.38it/s]
5/128 0.508G 1.34 2.17 0.222 1: 100% 16/16 [00:02<00:00, 5.51it/s]
6/128 0.508G 1.26 1.76 0.25 1: 100% 16/16 [00:04<00:00, 3.51it/s]
7/128 0.508G 1.32 1.3 0.306 1: 100% 16/16 [00:02<00:00, 6.76it/s]
8/128 0.508G 1.27 1.57 0.333 1: 100% 16/16 [00:02<00:00, 6.99it/s]
9/128 0.508G 1.38 1.5 0.306 1: 100% 16/16 [00:02<00:00, 6.51it/s]
10/128 0.508G 1.3 1.39 0.278 1: 100% 16/16 [00:02<00:00, 5.73it/s]
11/128 0.508G 1.3 1.55 0.361 1: 100% 16/16 [00:03<00:00, 4.95it/s]
12/128 0.508G 1.28 1.45 0.306 1: 100% 16/16 [00:02<00:00, 6.98it/s]
13/128 0.508G 1.28 1.33 0.528 1: 100% 16/16 [00:02<00:00, 6.34it/s]
14/128 0.508G 1.24 1.19 0.417 1: 100% 16/16 [00:02<00:00, 6.90it/s]
15/128 0.508G 1.27 1.81 0.222 1: 100% 16/16 [00:03<00:00, 4.79it/s]
16/128 0.508G 1.25 1.52 0.361 1: 100% 16/16 [00:02<00:00, 6.45it/s]
17/128 0.508G 1.28 1.2 0.361 1: 100% 16/16 [00:02<00:00, 6.15it/s]
18/128 0.508G 1.25 1.33 0.528 1: 100% 16/16 [00:02<00:00, 6.79it/s]
19/128 0.508G 1.18 1.17 0.5 1: 100% 16/16 [00:02<00:00, 6.67it/s]
20/128 0.508G 1.23 1.33 0.306 1: 100% 16/16 [00:04<00:00, 3.52it/s]
21/128 0.508G 1.21 1.39 0.417 1: 100% 16/16 [00:02<00:00, 6.89it/s]
22/128 0.508G 1.18 1.36 0.528 1: 100% 16/16 [00:02<00:00, 6.43it/s]
23/128 0.508G 1.14 1.38 0.5 1: 100% 16/16 [00:02<00:00, 6.70it/s]
24/128 0.508G 1.17 1.3 0.556 1: 100% 16/16 [00:03<00:00, 4.59it/s]
25/128 0.508G 1.2 1.13 0.583 1: 100% 16/16 [00:02<00:00, 6.26it/s]
26/128 0.508G 1.11 1.12 0.528 1: 100% 16/16 [00:02<00:00, 6.69it/s]
27/128 0.508G 1.12 1.06 0.583 1: 100% 16/16 [00:02<00:00, 6.37it/s]
28/128 0.508G 1.12 1.45 0.417 1: 100% 16/16 [00:02<00:00, 6.95it/s]
29/128 0.508G 1.19 1.11 0.5 1: 100% 16/16 [00:03<00:00, 4.33it/s]
30/128 0.508G 1.14 1.2 0.583 1: 100% 16/16 [00:02<00:00, 6.86it/s]
31/128 0.508G 1.1 1.34 0.5 1: 100% 16/16 [00:02<00:00, 5.83it/s]
32/128 0.508G 1.17 2.32 0.278 1: 100% 16/16 [00:02<00:00, 6.40it/s]
33/128 0.508G 1.11 1.02 0.667 1: 100% 16/16 [00:02<00:00, 5.47it/s]
34/128 0.508G 1.16 1.37 0.5 1: 100% 16/16 [00:03<00:00, 5.17it/s]
35/128 0.508G 1.1 1.12 0.472 1: 100% 16/16 [00:02<00:00, 6.79it/s]
36/128 0.508G 1.08 1.2 0.556 1: 100% 16/16 [00:03<00:00, 4.22it/s]
37/128 0.508G 1.11 1.08 0.556 1: 100% 16/16 [00:02<00:00, 6.21it/s]
38/128 0.508G 1.13 1.26 0.528 1: 100% 16/16 [00:03<00:00, 4.65it/s]
39/128 0.508G 1.12 1.11 0.667 1: 100% 16/16 [00:02<00:00, 6.73it/s]
40/128 0.508G 1.11 1.19 0.639 1: 100% 16/16 [00:02<00:00, 6.53it/s]
41/128 0.508G 1.07 0.947 0.556 1: 100% 16/16 [00:02<00:00, 6.87it/s]
42/128 0.508G 1.07 1.18 0.611 1: 100% 16/16 [00:03<00:00, 5.17it/s]
43/128 0.508G 1.14 1.44 0.528 1: 100% 16/16 [00:02<00:00, 5.41it/s]
44/128 0.508G 1.05 1.01 0.667 1: 100% 16/16 [00:02<00:00, 6.64it/s]
45/128 0.508G 1.08 1.14 0.639 1: 100% 16/16 [00:02<00:00, 6.77it/s]
46/128 0.508G 1.07 1.33 0.528 1: 100% 16/16 [00:02<00:00, 6.31it/s]
47/128 0.508G 1.03 1 0.639 1: 100% 16/16 [00:03<00:00, 4.78it/s]
48/128 0.508G 1.04 1.71 0.611 1: 100% 16/16 [00:02<00:00, 5.78it/s]
49/128 0.508G 1.04 1.64 0.528 1: 100% 16/16 [00:02<00:00, 6.66it/s]
50/128 0.508G 1.02 1 0.75 1: 100% 16/16 [00:02<00:00, 6.63it/s]
51/128 0.508G 1.02 1.11 0.667 1: 100% 16/16 [00:02<00:00, 6.63it/s]
52/128 0.508G 1.06 1.59 0.611 1: 100% 16/16 [00:03<00:00, 4.26it/s]
53/128 0.508G 0.973 1.07 0.667 1: 100% 16/16 [00:02<00:00, 6.46it/s]
54/128 0.508G 0.925 1.34 0.556 1: 100% 16/16 [00:02<00:00, 6.46it/s]
55/128 0.508G 1.1 0.927 0.667 1: 100% 16/16 [00:03<00:00, 4.46it/s]
56/128 0.508G 1 1.97 0.583 1: 100% 16/16 [00:05<00:00, 3.06it/s]
57/128 0.508G 0.993 1.34 0.611 1: 100% 16/16 [00:02<00:00, 6.75it/s]
58/128 0.508G 0.954 1.17 0.639 1: 100% 16/16 [00:02<00:00, 6.50it/s]
59/128 0.508G 1.03 1.54 0.5 1: 100% 16/16 [00:02<00:00, 6.59it/s]
60/128 0.508G 1.01 1.12 0.611 1: 100% 16/16 [00:03<00:00, 5.32it/s]
61/128 0.508G 1 1.13 0.583 1: 100% 16/16 [00:03<00:00, 5.28it/s]
62/128 0.508G 0.943 0.986 0.639 1: 100% 16/16 [00:02<00:00, 6.75it/s]
63/128 0.508G 0.909 1.12 0.639 1: 100% 16/16 [00:02<00:00, 6.97it/s]
64/128 0.508G 0.888 0.867 0.75 1: 100% 16/16 [00:02<00:00, 6.32it/s]
65/128 0.508G 0.958 0.975 0.667 1: 100% 16/16 [00:03<00:00, 4.41it/s]
66/128 0.508G 0.939 0.947 0.639 1: 100% 16/16 [00:02<00:00, 6.54it/s]
67/128 0.508G 1.02 1.11 0.694 1: 100% 16/16 [00:03<00:00, 5.04it/s]
68/128 0.508G 0.998 0.971 0.667 1: 100% 16/16 [00:02<00:00, 5.55it/s]
69/128 0.508G 0.968 0.98 0.694 1: 100% 16/16 [00:03<00:00, 4.52it/s]
70/128 0.508G 0.965 1.11 0.722 1: 100% 16/16 [00:02<00:00, 6.55it/s]
71/128 0.508G 0.965 1.47 0.583 1: 100% 16/16 [00:02<00:00, 6.84it/s]
72/128 0.508G 0.953 1.2 0.611 1: 100% 16/16 [00:02<00:00, 6.54it/s]
73/128 0.508G 0.863 0.772 0.722 1: 100% 16/16 [00:02<00:00, 6.90it/s]
74/128 0.508G 0.946 0.884 0.667 1: 100% 16/16 [00:03<00:00, 4.25it/s]
75/128 0.508G 0.911 0.942 0.694 1: 100% 16/16 [00:02<00:00, 6.78it/s]
76/128 0.508G 0.964 1.16 0.694 1: 100% 16/16 [00:02<00:00, 6.80it/s]
77/128 0.508G 0.917 1.2 0.694 1: 100% 16/16 [00:02<00:00, 6.44it/s]
78/128 0.508G 0.941 0.955 0.639 1: 100% 16/16 [00:02<00:00, 6.22it/s]
79/128 0.508G 0.885 1.02 0.722 1: 100% 16/16 [00:03<00:00, 4.58it/s]
80/128 0.508G 0.864 0.802 0.694 1: 100% 16/16 [00:02<00:00, 6.33it/s]
81/128 0.508G 0.908 1.11 0.833 1: 100% 16/16 [00:02<00:00, 6.52it/s]
82/128 0.508G 0.915 0.843 0.778 1: 100% 16/16 [00:02<00:00, 6.82it/s]
83/128 0.508G 0.899 1.14 0.722 1: 100% 16/16 [00:03<00:00, 4.96it/s]
84/128 0.508G 0.826 0.81 0.75 1: 100% 16/16 [00:02<00:00, 5.77it/s]
85/128 0.508G 0.831 0.883 0.694 1: 100% 16/16 [00:02<00:00, 6.61it/s]
86/128 0.508G 0.804 0.95 0.694 1: 100% 16/16 [00:02<00:00, 6.42it/s]
87/128 0.508G 0.805 0.916 0.694 1: 100% 16/16 [00:02<00:00, 6.60it/s]
88/128 0.508G 0.824 0.936 0.667 1: 100% 16/16 [00:03<00:00, 4.40it/s]
89/128 0.508G 0.854 0.854 0.639 1: 100% 16/16 [00:02<00:00, 6.48it/s]
90/128 0.508G 0.79 1.14 0.694 1: 100% 16/16 [00:02<00:00, 6.72it/s]
91/128 0.508G 0.83 0.848 0.75 1: 100% 16/16 [00:02<00:00, 6.59it/s]
92/128 0.508G 0.805 1.32 0.639 1: 100% 16/16 [00:02<00:00, 6.47it/s]
93/128 0.508G 0.813 1.22 0.75 1: 100% 16/16 [00:03<00:00, 4.23it/s]
94/128 0.508G 0.796 0.91 0.722 1: 100% 16/16 [00:02<00:00, 6.68it/s]
95/128 0.508G 0.823 0.778 0.75 1: 100% 16/16 [00:02<00:00, 6.70it/s]
96/128 0.508G 0.827 0.898 0.806 1: 100% 16/16 [00:02<00:00, 6.50it/s]
97/128 0.508G 0.777 0.833 0.778 1: 100% 16/16 [00:02<00:00, 5.78it/s]
98/128 0.508G 0.79 0.735 0.806 1: 100% 16/16 [00:03<00:00, 4.78it/s]
99/128 0.508G 0.824 0.797 0.778 1: 100% 16/16 [00:02<00:00, 6.19it/s]
100/128 0.508G 0.802 0.893 0.806 1: 100% 16/16 [00:02<00:00, 5.94it/s]
101/128 0.508G 0.778 1.11 0.778 1: 100% 16/16 [00:02<00:00, 6.61it/s]
102/128 0.508G 0.795 1.15 0.722 1: 100% 16/16 [00:03<00:00, 4.30it/s]
103/128 0.508G 0.777 1.54 0.667 1: 100% 16/16 [00:02<00:00, 6.39it/s]
104/128 0.508G 0.764 0.916 0.722 1: 100% 16/16 [00:02<00:00, 6.66it/s]
105/128 0.508G 0.737 1.04 0.778 1: 100% 16/16 [00:02<00:00, 6.57it/s]
106/128 0.508G 0.689 0.792 0.75 1: 100% 16/16 [00:02<00:00, 6.55it/s]
107/128 0.508G 0.769 0.945 0.75 1: 100% 16/16 [00:03<00:00, 4.40it/s]
108/128 0.508G 0.78 1.21 0.75 1: 100% 16/16 [00:02<00:00, 6.61it/s]
109/128 0.508G 0.768 0.958 0.75 1: 100% 16/16 [00:02<00:00, 6.37it/s]
110/128 0.508G 0.802 0.953 0.75 1: 100% 16/16 [00:02<00:00, 6.41it/s]
111/128 0.508G 0.765 0.71 0.75 1: 100% 16/16 [00:02<00:00, 5.42it/s]
112/128 0.508G 0.709 1.07 0.722 1: 100% 16/16 [00:03<00:00, 5.15it/s]
113/128 0.508G 0.683 1.1 0.694 1: 100% 16/16 [00:02<00:00, 6.57it/s]
114/128 0.508G 0.685 0.892 0.778 1: 100% 16/16 [00:02<00:00, 6.41it/s]
115/128 0.508G 0.678 0.78 0.722 1: 100% 16/16 [00:02<00:00, 6.25it/s]
116/128 0.508G 0.714 1.19 0.722 1: 100% 16/16 [00:03<00:00, 4.29it/s]
117/128 0.508G 0.718 0.777 0.694 1: 100% 16/16 [00:02<00:00, 6.04it/s]
118/128 0.508G 0.744 0.855 0.778 1: 100% 16/16 [00:02<00:00, 6.72it/s]
119/128 0.508G 0.732 0.708 0.75 1: 100% 16/16 [00:02<00:00, 6.66it/s]
120/128 0.508G 0.7 0.88 0.778 1: 100% 16/16 [00:02<00:00, 5.85it/s]
121/128 0.508G 0.687 0.852 0.778 1: 100% 16/16 [00:03<00:00, 4.65it/s]
122/128 0.508G 0.671 1.01 0.778 1: 100% 16/16 [00:02<00:00, 6.46it/s]
123/128 0.508G 0.695 0.708 0.75 1: 100% 16/16 [00:02<00:00, 6.40it/s]
124/128 0.508G 0.685 0.725 0.778 1: 100% 16/16 [00:02<00:00, 6.69it/s]
125/128 0.508G 0.681 0.991 0.75 1: 100% 16/16 [00:03<00:00, 4.79it/s]
126/128 0.508G 0.674 0.72 0.75 1: 100% 16/16 [00:03<00:00, 4.96it/s]
127/128 0.508G 0.674 0.733 0.75 1: 100% 16/16 [00:02<00:00, 6.52it/s]
128/128 0.508G 0.687 0.682 0.75 1: 100% 16/16 [00:02<00:00, 6.48it/s]

Training complete (0.105 hours)
Results saved to runs/train-cls/exp
Predict: python classify/predict.py --weights runs/train-cls/exp/weights/best.pt --source im.jpg
Validate: python classify/val.py --weights runs/train-cls/exp/weights/best.pt --data Banco-Imagem-1
Export: python export.py --weights runs/train-cls/exp/weights/best.pt --include onnx
PyTorch Hub: model = torch.hub.load('ultralytics/yolov5', 'custom', 'runs/train-cls/exp/weights/best.pt')
Visualize: https://netron.app

CPU times: user 4.67 s, sys: 452 ms, total: 5.12 s
Wall time: 6min 43s

!python classify/val.py --weights runs/train-cls/exp/weights/best.pt --data $DATASET_NAME

classify/val: data=Banco-Imagem-1, weights=['runs/train-cls/exp/weights/best.pt'], batch_size=128, imgsz=224, device=, workers=8, verbose=True, project=runs/val-cls, name=exp, exist_ok=False, half=False, dnn=False
YOLOv5 🚀 v7.0-230-g53efd07 Python-3.10.12 torch-2.1.0+cu118 CUDA:0 (Tesla T4, 15102MiB)

Fusing layers...
Model summary: 117 layers, 1214869 parameters, 0 gradients, 2.9 GFLOPs
testing: 100% 1/1 [00:00<00:00, 1.05it/s]
Class Images top1_acc top5_acc
all 36 0.639 1
avioes 7 0.571 1
barcos 6 0.667 1
carros 11 0.545 1
helicopteros 8 0.875 1
motos 4 0.5 1
Speed: 0.1ms pre-process, 14.8ms inference, 0.6ms post-process per image at shape (1, 3, 224, 224)
Results saved to runs/val-cls/exp

```
</details>

### Evidências do treinamento

#### Gráficos de precisão e perdas


#### Matriz de Confusão


#### Inferindo com o modelo personalizado
```
#Pega a localização de uma imagem do conjunto de testes ou validações
if os.path.exists(os.path.join(dataset.location, "test")):
split_path = os.path.join(dataset.location, "test")
else:
os.path.join(dataset.location, "valid")
example_class = os.listdir(split_path)[4]
example_image_name = os.listdir(os.path.join(split_path, example_class))[4]
example_image_path = os.path.join(split_path, example_class, example_image_name)
os.environ["TEST_IMAGE_PATH"] = example_image_path

print(f"Inferindo sobre um exemplo da classe '{example_class}'")

#Infer
!python classify/predict.py --weights runs/train-cls/exp/weights/best.pt --source $TEST_IMAGE_PATH

Inferindo sobre um exemplo da classe 'carros'
classify/predict: weights=['runs/train-cls/exp/weights/best.pt'], source=/content/yolov5/Banco-Imagem-1/test/carros/00012_jpg.rf.9f0d32646e83139878c5788b040038f7.jpg, data=data/coco128.yaml, imgsz=[224, 224], device=, view_img=False, save_txt=False, nosave=False, augment=False, visualize=False, update=False, project=runs/predict-cls, name=exp, exist_ok=False, half=False, dnn=False, vid_stride=1
YOLOv5 🚀 v7.0-230-g53efd07 Python-3.10.12 torch-2.1.0+cu118 CUDA:0 (Tesla T4, 15102MiB)

Fusing layers...
Model summary: 117 layers, 1214869 parameters, 0 gradients, 2.9 GFLOPs
image 1/1 /content/yolov5/Banco-Imagem-1/test/carros/00012_jpg.rf.9f0d32646e83139878c5788b040038f7.jpg: 224x224 carros 0.91, avioes 0.08, motos 0.01, helicopteros 0.00, barcos 0.00, 2.7ms
Speed: 0.3ms pre-process, 2.7ms inference, 5.1ms NMS per image at shape (1, 3, 224, 224)
Results saved to runs/predict-cls/exp14
```

```
#### Modelo treinado com 80% ou mais de acurácia/precisão
=========================================================
```

```
#carro
import requests
image_url = "https://i.imgur.com/GB9Tihf.jpg"
response = requests.get(image_url)
response.raise_for_status()
with open('carro.jpg', 'wb') as handler:
handler.write(response.content)

!python classify/predict.py --weights ./weights/yolov5x-cls.pt --source carro.jpg

classify/predict: weights=['./weigths/yolov5x-cls.pt'], source=carro.jpg, data=data/coco128.yaml, imgsz=[224, 224], device=, view_img=False, save_txt=False, nosave=False, augment=False, visualize=False, update=False, project=runs/predict-cls, name=exp, exist_ok=False, half=False, dnn=False, vid_stride=1
YOLOv5 🚀 v7.0-230-g53efd07 Python-3.10.12 torch-2.1.0+cu118 CUDA:0 (Tesla T4, 15102MiB)

Fusing layers...
Model summary: 264 layers, 48072600 parameters, 0 gradients, 129.9 GFLOPs
image 1/1 /content/yolov5/carro.jpg: 224x224 sports car 0.95, race car 0.02, convertible 0.01, car wheel 0.00, grille 0.00, 12.9ms
Speed: 0.4ms pre-process, 12.9ms inference, 6.9ms NMS per image at shape (1, 3, 224, 224)
Results saved to runs/predict-cls/exp13

### Modelo treinado com ao menos 50% de acurácia/precisão
=========================================================
```
```

#Moto
import requests
image_url = "https://i.imgur.com/ASwjAT5.jpg"
response = requests.get(image_url)
response.raise_for_status()
with open('moto.jpg', 'wb') as handler:
handler.write(response.content)

!python classify/predict.py --weights ./weights/yolov5m-cls.pt --source moto.jpg

classify/predict: weights=['./weigths/yolov5m-cls.pt'], source=moto.jpg, data=data/coco128.yaml, imgsz=[224, 224], device=, view_img=False, save_txt=False, nosave=False, augment=False, visualize=False, update=False, project=runs/predict-cls, name=exp, exist_ok=False, half=False, dnn=False, vid_stride=1
YOLOv5 🚀 v7.0-230-g53efd07 Python-3.10.12 torch-2.1.0+cu118 CUDA:0 (Tesla T4, 15102MiB)

Fusing layers...
Model summary: 166 layers, 12947192 parameters, 0 gradients, 31.7 GFLOPs
image 1/1 /content/yolov5/moto.jpg: 224x224 moped 0.64, scooter 0.17, disc brake 0.06, crash helmet 0.05, snowmobile 0.01, 5.4ms
Speed: 0.4ms pre-process, 5.4ms inference, 6.9ms NMS per image at shape (1, 3, 224, 224)
Results saved to runs/predict-cls/exp16

```

## Roboflow

Banco-Imagem > 2023-10-24 9:29pm

https://universe.roboflow.com/eniokilder/banco-imagem

Provided by a Roboflow user
License: CC BY 4.0

## HuggingFace

Link para o HuggingFace:

https://huggingface.co/datasets/eniokilder/Banco-Imagem

原始信息汇总

数据集概述

项目类型

• 类型: 物体分类
• 模型: YOLOv5
• 编程语言: PyTorch

性能

• 模型训练性能: 98.6%

训练细节

• 数据集名称: Banco-Imagem-1
• 类别数: 5
• 训练周期: 128
• 批次大小: 16
• 图像尺寸: 320x320
• 预训练权重: yolov5n-cls.pt
• 优化器: Adam
• 初始学习率: 0.001
• 权重衰减: 5e-05
• 标签平滑: 0.1

训练输出

• 训练时间: 0.105小时
• 结果保存路径: runs/train-cls/exp

验证结果

• 验证数据集: Banco-Imagem-1
• 验证批次大小: 128
• 图像尺寸: 224x224
• 验证结果:
  • 总体准确率: 0.639
  • 各类别准确率:
    • avioes: 0.571
    • barcos: 0.667
    • carros: 0.545
    • helicopteros: 0.875
    • motos: 0.5

预测示例

• 示例类别: carros
• 预测结果: carros 0.91, avioes 0.08, motos 0.01, helicopteros 0.00, barcos 0.00

搜集汇总

背景与挑战

背景概述

该数据集是一个小型图像分类数据集，包含370张图像，用于训练和验证YOLOv5模型，模型训练后达到了98.6%的高准确率。