zhanghang1989/PyTorch-Encoding

A CV toolkit for my papers.

What it solves

PyTorch-Encoding provides a library of specialized encoding modules and pre-trained models for computer vision tasks, specifically focusing on image classification and semantic segmentation.

How it works

The project implements various encoding techniques and network architectures, including Split-Attention Networks (ResNeSt) and Texture Encoding Networks (Deep TEN), to improve how visual information is captured and cross-referenced in neural networks.

Who it’s for

Researchers and developers working with PyTorch who need high-performance vision models for classification and segmentation tasks.

Highlights

  • own model zoo containing image classification and semantic segmentation models.
  • Implementation of ResNeSt (Split-Attention Networks).
  • Implementation of Deep TEN (Texture Encoding Network).
  • Implementation of Context Encoding for semantic segmentation.

Related

  • Project
    opendatalab/OmniDocBench

    A comprehensive benchmark for evaluating document parsing in real-world scenarios, featuring rich annotations for layout detection, OCR, table, and formula recognition.

  • Project
    apple-aiml-research/ml-fastvit

    A fast hybrid vision transformer that uses structural reparameterization to achieve low-latency image classification on mobile devices.

  • Project
    apple-aiml-research/ml-hypersim

    Hypersim is a large synthetic indoor dataset (≈77 k HDR images, 1.9 TB) with per‑pixel geometry, semantics, and material channels, plus a V‑Ray‑based toolkit for generating and editing similar data. It’s designed for training and evaluating scene‑understanding models (depth, normals, segmentation, intrinsic image decomposition) and includes a ready‑made train/val/test split.

  • Project
    bytedance/Protenix

    An open-source biomolecular structure prediction framework that provides high-accuracy 3D modeling of proteins and RNA, designed to be an accessible alternative to AlphaFold3.