albumentations-teamOfficial·10 Agent Skills Included

benchmark

Benchmark image and video augmentation library performance

Measures throughput of image and video augmentation libraries like AlbumentationsX, torchvision, Kornia, Pillow, and DALI across CPU and GPU. Runs micro and DataLoader benchmarks on local machines or detached Google Cloud VMs with adaptive warmup and slow-transform early stopping. Generates comparison tables, speedup plots, and publication-ready figures without manual spreadsheet work.

95|3

↗

npx skills add albumentations-team/benchmark --all -g -y

All Skills in This Repository (10)

Pure Emerald Level Indicators

📦 In Repo

albumentations-team

paper-benchmark-execution

Orchestrate paper benchmark execution for RGB, multichannel, DataLoader, and video benchmarks.

benchmark-results-publisher

Convert benchmark result JSONs into documentation-ready tables and figures.

gcp-benchmark-triage

Triage detached GCP benchmark runs by inspecting logs and validating artifacts.

documentation-generator

Automate benchmark README table, speedup plot, and metadata updates.

pre-commit-validation

Automate pre-commit checks and auto-fixes for Python projects using Ruff, Mypy, Pyright, and ShellCheck.

performance-analysis

Analyze benchmark result JSON to detect slow transforms, warmup instability, and regressions.

benchmark-runner

Automate image and video augmentation benchmarks across libraries with YAML configs.

transform-spec-validator

Validate transform specification files for the benchmark runner.

library-integration

Implement image/video augmentation library transforms and register matrix specs.

paper-coverage-validator

Validate benchmark artifacts for required RGB micro and RGB DataLoader sections.

Official

Intermediate

Frequently Asked Questions

FAQPage Schema

How to install albumentations benchmark?▼

Run `npx skills add albumentations-team/benchmark --all -g -y` in your terminal to install all skills in this suite globally.

Which augmentation library is fastest?▼

This suite measures throughput for AlbumentationsX, torchvision, Kornia, Pillow, and DALI across micro and DataLoader regimes so you can compare real numbers instead of guessing.

Can I run benchmarks on Google Cloud?▼

Yes. The suite launches detached GCP VMs, stages dataset tarballs, runs the benchmark, uploads results to GCS, and cleans up the VM automatically.

How do I add a new augmentation library to the benchmark?▼

The library-integration skill walks you through creating a transform spec file, registering it in the benchmark matrix, and generating baseline results.

Do I need coding experience to run benchmarks?▼

No. Your agent uses the checked-in YAML configs and skills to plan, run, validate, and publish benchmarks from plain-English requests.