{"categories":[{"label":"Artificial Intelligence","url":"https://skillfed.io/packages/category/scientific-engineering-artificial-intelligence/4"}],"enrichment":{"capability":"Provides PyTorch-based neural network models for image semantic segmentation, with 12 encoder-decoder architectures and 800+ pretrained encoders, usable in two lines of code.","skillfed_tags":["computer-vision","semantic-segmentation","pretrained-models"],"use_cases":["Train a binary segmentation model on custom image data (e.g., object boundaries) by choosing an encoder and decoder, then fine-tuning with your labeled dataset.","Load a pretrained Segformer or DPT model for immediate inference on new images without training.","Export a trained segmentation model to ONNX format for deployment in non-Python environments or edge devices.","Experiment with different encoder-decoder combinations to find the best accuracy-speed tradeoff for your segmentation task.","Build a multiclass segmentation pipeline that assigns multiple semantic labels to different regions of an image."],"what_it_does":"Segmentation Models PyTorch (SMP) is a library that wraps semantic segmentation architectures on top of PyTorch, combining encoder-decoder pairs to produce pixel-level classification masks. It abstracts away the boilerplate of building segmentation models by providing 12 pre-built architectures (Unet, Unet++, Segformer, DPT, DeepLabV3+, and others) that can be instantiated with a choice of encoder backbone and pretrained weights in minimal code.\n\nThe library's main value is its breadth of pretrained encoders\u2014800+ convolution and transformer-based models sourced from timm and huggingface-hub\u2014which you can mix and match with any decoder architecture. It includes training utilities (Dice, Jaccard, Tversky losses and metrics), supports ONNX export and torch.compile, and integrates with huggingface-hub for model sharing. The typical workflow is to instantiate a model, optionally load pretrained encoder weights, configure data preprocessing to match the encoder's training regime, then train or fine-tune on your segmentation task.","worth_installing":"Yes. The library is actively maintained, has low install friction, carries no known vulnerabilities, and offers a well-designed API for a common computer vision task. Install it if you need to build or fine-tune semantic segmentation models; the breadth of pretrained encoders and multiple architectures make it a practical choice over building from scratch. The MIT license poses no restrictions."},"id":"segmentation-models-pytorch","links":{"html":"https://skillfed.io/packages/segmentation-models-pytorch","md":"https://skillfed.io/packages/segmentation-models-pytorch.md","pypi":"https://pypi.org/project/segmentation-models-pytorch/"},"maintenance":{"status":"active"},"meta":{"latest_release":"2025-04-17","license_spdx":null,"license_treatment":"permissive","name":"segmentation-models-pytorch","python_support":"supports_current","summary":"Image segmentation models with pre-trained backbones. PyTorch."},"popularity":{"monthly_downloads":692350,"position":5321,"tier":"top_15000"},"security":{"n_vulnerabilities":0},"version":"0.5.0"}