Python library with Neural Networks for Image
Segmentation based on PyTorch.

The main features of this library are:

• High level API (just two lines to create neural network)
• 5 models architectures for binary and multi class segmentation (including legendary Unet)
• 46 available encoders for each architecture
• All encoders have pre-trained weights for faster and better convergence

Table of content

1. Quick start
2. Examples
3. Models
   1. Architectures
   2. Encoders
4. Models API
   1. Input channels
   2. Auxiliary classification output
   3. Depth
5. Installation
6. Competitions won with the library
7. License
8. Contributing

Quick start

Since the library is built on the PyTorch framework, created segmentation model is just a PyTorch nn.Module, which can be created as easy as:

import segmentation_models_pytorch as smp

model = smp.Unet()

Depending on the task, you can change the network architecture by choosing backbones with fewer or more parameters and use pretrainded weights to initialize it:

model = smp.Unet('resnet34', encoder_weights='imagenet')

Change number of output classes in the model:

model = smp.Unet('resnet34', classes=3, activation='softmax')

All models have pretrained encoders, so you have to prepare your data the same way as during weights pretraining:

from segmentation_models_pytorch.encoders import get_preprocessing_fn

preprocess_input = get_preprocessing_fn('resnet18', pretrained='imagenet')

Examples

• Training model for cars segmentation on CamVid dataset here.
• Training SMP model with Catalyst (high-level framework for PyTorch), Ttach (TTA library for PyTorch) and Albumentations (fast image augmentation library) - here

Models

Architectures

• Unet
• Linknet
• FPN
• PSPNet
• PAN

Encoders

Encoder	Weights	Params, M
resnet18	imagenet	11M
resnet34	imagenet	21M
resnet50	imagenet	23M
resnet101	imagenet	42M
resnet152	imagenet	58M
resnext50_32x4d	imagenet	22M
resnext101_32x8d	imagenet instagram	86M
resnext101_32x16d	instagram	191M
resnext101_32x32d	instagram	466M
resnext101_32x48d	instagram	826M
dpn68	imagenet	11M
dpn68b	imagenet+5k	11M
dpn92	imagenet+5k	34M
dpn98	imagenet	58M
dpn107	imagenet+5k	84M
dpn131	imagenet	76M
vgg11	imagenet	9M
vgg11_bn	imagenet	9M
vgg13	imagenet	9M
vgg13_bn	imagenet	9M
vgg16	imagenet	14M
vgg16_bn	imagenet	14M
vgg19	imagenet	20M
vgg19_bn	imagenet	20M
senet154	imagenet	113M
se_resnet50	imagenet	26M
se_resnet101	imagenet	47M
se_resnet152	imagenet	64M
se_resnext50_32x4d	imagenet	25M
se_resnext101_32x4d	imagenet	46M
densenet121	imagenet	6M
densenet169	imagenet	12M
densenet201	imagenet	18M
densenet161	imagenet	26M
inceptionresnetv2	imagenet imagenet+background	54M
inceptionv4	imagenet imagenet+background	41M
efficientnet-b0	imagenet	4M
efficientnet-b1	imagenet	6M
efficientnet-b2	imagenet	7M
efficientnet-b3	imagenet	10M
efficientnet-b4	imagenet	17M
efficientnet-b5	imagenet	28M
efficientnet-b6	imagenet	40M
efficientnet-b7	imagenet	63M
mobilenet_v2	imagenet	2M
xception	imagenet	22M

Models API

• model.encoder - pretrained backbone to extract features of different spatial resolution
• model.decoder - depends on models architecture (Unet/Linknet/PSPNet/FPN)
• model.segmentation_head - last block to produce required number of mask channels (include also optional upsampling and activation)
• model.classification_head - optional block which create classification head on top of encoder
• model.forward(x) - sequentially pass x through model`s encoder, decoder and segmentation head (and classification head if specified)

Input channels

Input channels parameter allow you to create models, which process tensors with arbitrary number of channels. If you use pretrained weights from imagenet - weights of first convolution will be reused for 1- or 2- channels inputs, for input channels > 4 weights of first convolution will be initialized randomly.

model = smp.FPN('resnet34', in_channels=1)
mask = model(torch.ones([1, 1, 64, 64]))

Auxiliary classification output

All models support aux_params parameters, which is default set to None. If aux_params = None than classification auxiliary output is not created, else model produce not only mask, but also label output with shape NC. Classification head consist of GlobalPooling->Dropout(optional)->Linear->Activation(optional) layers, which can be configured by aux_params as follows:

aux_params=dict(
    pooling='avg',             # one of 'avg', 'max'
    dropout=0.5,               # dropout ratio, default is None
    activation='sigmoid',      # activation function, default is None
    classes=4,                 # define number of output labels
)
model = smp.Unet('resnet34', classes=4, aux_params=aux_params)
mask, label = model(x)

Depth

Depth parameter specify a number of downsampling operations in encoder, so you can make your model lighted if specify smaller depth.

model = smp.Unet('resnet34', encoder_depth=4)

Installation

PyPI version:

$ pip install segmentation-models-pytorch

Latest version from source:

$ pip install git+https://github.com/qubvel/segmentation_models.pytorch

Competitions won with the library

Segmentation Models package is widely used in the image segmentation competitions. Here you can find competitions, names of the winners and links to their solutions.

License

Project is distributed under MIT License

Contributing

Run test

$ docker build -f docker/Dockerfile.dev -t smp:dev . && docker run --rm smp:dev pytest -p no:cacheprovider

Generate table

$ docker build -f docker/Dockerfile.dev -t smp:dev . && docker run --rm smp:dev python misc/generate_table.py