pytorch-semseg and pytorch-segmentation
These are competitors offering overlapping implementations of semantic segmentation architectures in PyTorch, with the first providing a broader collection of classical architectures while the second emphasizes integrated datasets and loss functions alongside model implementations.
About pytorch-semseg
meetps/pytorch-semseg
Semantic Segmentation Architectures Implemented in PyTorch
Implements 8+ segmentation architectures (PSPNet, FCN, U-Net, SegNet, LinkNet, ICNet, FRRN) with pretrained weights and integrated support for 7 major datasets (Pascal VOC, Cityscapes, ADE20K, CamVid, NYUDv2, etc.). Provides YAML-based configuration for training with modular components including multiple optimizers, learning rate schedules, data augmentation pipelines, and loss functions. Includes utilities for validation with flip augmentation, inference on custom images with optional DenseCRF post-processing, and real-time FPS measurement.
About pytorch-segmentation
yassouali/pytorch-segmentation
:art: Semantic segmentation models, datasets and losses implemented in PyTorch.
Implements multiple encoder-decoder architectures (DeepLab V3+, PSPNet, UperNet, U-Net, etc.) with atrous convolution and multi-scale parsing strategies for dense pixel-level prediction. Provides specialized loss functions including Lovász-Softmax for direct mIoU optimization and focal loss for handling class imbalance, alongside poly and one-cycle learning rate schedulers commonly used in segmentation workflows. Supports Pascal VOC, Cityscapes, ADE20K, and COCO Stuff datasets with JSON-based configuration for reproducible training pipelines.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work