opennn-pytorch 1.0.4

Neural Networks library for image classification task.

Open Neural Networks library for image classification.

Table of content

1. Quick start
2. Warnings
3. Encoders
4. Decoders
5. Pretrained
6. Datasets
7. Losses
8. Metrics
9. Optimizers
10. Schedulers
11. Examples

Quick start

1. Straight install.

1.1 Install torch with cuda.

pip install -U torch --extra-index-url https://download.pytorch.org/whl/cu113

1.2 Install opennn_pytorch.

pip install -U opennn_pytorch

2. Dockerfile.

cd docker/
docker build -t opennn:latest .

Warnings

1. Cuda is only supported for nvidia graphics cards.
2. Alexnet decoder doesn't support bce losses family.
3. Sometimes combination of dataset/encoder/decoder/loss can give bad results, try to combine others.
4. Custom cross-entropy support only mode when preds have (n, c) shape and labels have (n) shape.
5. Not all options in transform.yaml and config.yaml are required.
6. Mean and std in datasets section must be used in transform.yaml, for example [mean=[0.2859], std=[0.3530]] -> normalize: [[0.2859], [0.3530]]

Encoders

• LeNet [paper] [code]
• AlexNet [paper] [code]
• GoogleNet [paper] [code]
• ResNet18 [paper] [code]
• ResNet34 [paper] [code]
• ResNet50 [paper] [code]
• ResNet101 [paper] [code]
• ResNet152 [paper] [code]
• MobileNet [paper] [code]
• VGG-11 [paper] [code]
• VGG-16 [paper] [code]
• VGG-19 [paper] [code]

Decoders

• LeNet [paper] [code]
• AlexNet [paper] [code]
• Linear [docs] [code]

Pretrained

LeNet

Encoder	Decoder	Dataset	Weights	Configs	Logs
LeNet	LeNet	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL
LeNet	LeNet	FASHION-MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL
LeNet	Linear	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL
LeNet	Linear	FASHION-MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL

GoogleNet

Encoder	Decoder	Dataset	Weights	Configs	Logs
GoogleNet	Linear	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL

ResNet

Encoder	Decoder	Dataset	Weights	Configs	Logs
ResNet18	Linear	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL
ResNet34	Linear	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL

MobileNet

Encoder	Decoder	Dataset	Weights	Configs	Logs
MobileNet	Linear	MNIST	BEST, PLAN	CONFIG, TRANSFORM	TRAINVAL

Datasets

• MNIST [files] [code] [classes=10] [mean=[0.1307], std=[0.3801]]
• FASHION-MNIST [files] [code] [classes=10] [mean=[0.2859], std=[0.3530]]
• CIFAR-10 [files] [code] [classes=10] [mean=[0.491, 0.482, 0.446], std=[0.247, 0.243, 0.261]]
• CIFAR-100 [files] [code] [classes=100] [mean=[0.5071, 0.4867, 0.4408], std=[0.2675, 0.2565, 0.2761]]
• GTSRB [files] [code] [classes=43] [mean=unknown, std=unknown]
• CUSTOM [docs] [code] [example] [classes=nc] [mean=unknown, std=unknown]

Losses

• Cross-Entropy [pytorch, custom] [docs] [code]
• Binary-Cross-Entropy [pytorch, custom] [docs] [code]
• Binary-Cross-Entropy-With-Logits [pytorch, custom] [docs] [code]
• Mean-Squared-Error [pytorch, custom] [docs] [code]
• Mean-Absolute-Error [pytorch, custom] [docs] [code]

Metrics

• Accuracy [custom] [docs] [code]
• Precision [sklearn] [docs] [code]
• Recall [sklearn] [docs] [code]
• F1 [sklearn] [docs] [code]

Optimizers

• Adam [pytorch] [docs] [code]
• AdamW [pytorch] [docs] [code]
• Adamax [pytorch] [docs] [code]
• RAdam [pytorch] [docs] [code]
• NAdam [pytorch] [docs] [code]

Schedulers

• StepLR [pytorch] [docs] [code]
• MultiStepLR [pytorch] [docs] [code]

Examples

1. Run from yaml configs.

import opennn_pytorch
  
config = 'path to yaml config'  # check configs folder
opennn_pytorch.run(config)

2. Get encoder and decoder.

import opennn_pytorch
  
encoder_name = 'resnet18'
decoder_name = 'alexnet'
decoder_mode = 'decoder'
input_channels = 1
number_classes = 10
device = 'cuda'

encoder = opennn_pytorch.encoders.get_encoder(encoder_name, input_channels).to(device)
model = opennn_pytorch.decoders.get_decoder(decoder_name, encoder, number_classes, decoder_mode, device).to(device)

3.1 Get dataset.

import opennn_pytorch
from torchvision import transforms

transform_config = 'path to transform yaml config'
dataset_name = 'mnist'
datafiles = None
train_part = 0.7
valid_part = 0.2

transform_lst = opennn_pytorch.transforms_lst(transform_config)
transform = transforms.Compose(transform_lst)
  
train_data, valid_data, test_data = opennn_pytorch.datasets.get_dataset(dataset_name, train_part, valid_part, transform, datafiles)

3.2 Get custom dataset.

import opennn_pytorch
from torchvision import transforms

transform_config = 'path to transform yaml config'
dataset_name = 'custom'
images = 'path to folder with images'
annotation = 'path to annotation yaml file with image: class structure'
datafiles = (images, annotation)
train_part = 0.7
valid_part = 0.2

transform_lst = opennn_pytorch.transforms_lst(transform_config)
transform = transforms.Compose(transform_lst)
  
train_data, valid_data, test_data = opennn_pytorch.datasets.get_dataset(dataset_name, train_part, valid_part, transform, datafiles)

4. Get optimizer.

import opennn_pytorch

optim_name = 'adam'
lr = 1e-3
betas = (0.9, 0.999)
eps = 1e-8
weight_decay = 1e-6
optimizer = opennn_pytorch.optimizers.get_optimizer(optim_name, model, lr=lr, betas=betas, eps=opt_eps, weight_decay=weight_decay)

5. Get scheduler.

import opennn_pytorch

scheduler_name = 'steplr'
step = 10
gamma = 0.5
scheduler = opennn_pytorch.schedulers.get_scheduler(sched, optimizer, step=step, gamma=gamma, milestones=None)

6. Get loss function.

import opennn_pytorch

loss_fn = 'custom_mse'
loss_fn, one_hot = opennn_pytorch.losses.get_loss(loss_fn)

7. Get metrics functions.

import opennn_pytorch

metrics_names = ['accuracy', 'precision', 'recall', 'f1_score']
number_classes = 10
metrics_fn = opennn_pytorch.metrics.get_metrics(metrics_names, nc=number_classes)

8. Train/Test.

import opennn_pytorch

algorithm = 'train'
batch_size = 16
class_names = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
number_classes = 10
save_every = 5
epochs = 20

train_dataloader = torch.utils.data.DataLoader(train_data, batch_size=batch_size, shuffle=True)
valid_dataloader = torch.utils.data.DataLoader(valid_data, batch_size=batch_size, shuffle=False)
test_dataloader = torch.utils.data.DataLoader(test_data, batch_size=1, shuffle=False)

if algorithm == 'train':
  opennn_pytorch.algo.train(train_dataloader, valid_dataloader, model, optimizer, scheduler, loss_fn, metrics_fn, epochs, checkpoints, logs, device, save_every, one_hot, number_classes)
elif algorithm == 'test':
  test_logs = opennn_pytorch.algo.test(test_dataloader, model, loss_fn, metrics_fn, logs, device, one_hot, number_classes)
  if viz:
    os.mkdir(test_logs + '/vizualize', 0o777)
    for i in range(10):
      os.mkdir(test_logs + f'/vizualize/{i}', 0o777)
      opennn_pytorch.algo.vizualize(valid_data, model, device, {i: class_names[i] for i in range(number_classes)}, test_logs + f'/vizualize/{i}')

Citation

Project citation.

License

Project is distributed under MIT License.