# BloodAxe/pytorch-toolbelt

PyTorch extensions for fast R&D prototyping and Kaggle farming

Repository: https://github.com/BloodAxe/pytorch-toolbelt
Canonical: https://ross.abutalabs.com/products/pytorch-toolbelt
Language: Python
License: MIT
License Family: permissive
Topics: pytorch, kaggle, image-classification, image-segmentation, deep-learning, segmentation, python, image-processing, machine-learning, focal-loss, jaccard-loss, tta, test-time-augmentation, augmentation, object-detection, pipeline
Last push: 2025-10-09T19:41:20+00:00

## Health v2 (maintenance only)
Score: 44/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 46, release rhythm 8, longevity 100
- inputs: {"age_days": 2728, "days_push": 328, "days_rel": 650, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1574, forks 126 (observed 2026-08-28T04:05:05.688056+00:00)

## What it is
A Python library of PyTorch extensions providing building blocks for fast R&D prototyping, including encoder-decoder architectures, specialized losses, and GPU-friendly test-time augmentation. It is designed to complement high-level frameworks like Catalyst, Ignite, or fast.ai rather than replace them.

## Use cases
- build U-Net style encoder-decoder models for image segmentation
- apply focal, Dice, Jaccard, or Lovasz losses to segmentation training
- run test-time augmentation on GPU for classification or segmentation
- run inference on very large images like 5000x5000 tiles
- fix and restore random seeds for reproducible experiments
- prototype Kaggle competition pipelines quickly

## When to choose
- you already use PyTorch and want reusable modules, losses, and TTA utilities
- you compete in Kaggle vision competitions and want battle-tested building blocks
- you need GPU-friendly inference on huge images

## When to avoid
- you want a full high-level training framework like Catalyst, Ignite, or fast.ai
- your project does not use PyTorch
- you need non-vision deep learning tooling

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, image-processing, computer-vision, data-science
- domain: deep-learning, computer-vision, machine-learning, image-processing
- platform: python
- tags: pytorch, kaggle, segmentation, losses, test-time-augmentation, encoder-decoder, unet

## Member repositories
- BloodAxe/pytorch-toolbelt (main) score 44

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:05.688056+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T03:57:37.554684+00:00, confidence not recorded.
  - readme: https://github.com/BloodAxe/pytorch-toolbelt (fetched 2026-08-28T04:05:05.688056+00:00, sha 6c251efa419c)
  - registry_pypi: https://pypi.org/pypi/pytorch-toolbelt/json (fetched 2026-08-29T11:27:50.521506+00:00, sha 617a15a93e92)
- Data as of 2026-08-30T08:39:29.467469+00:00.