SkillA skill gives Relay a specialist method it can reuse on your work.AI & Agent ToolingCommunity
Medium riskSecurity review passed

huggingface-vision-trainer

Trains and fine-tunes vision models for object detection.

How it works

What huggingface-vision-trainer does

D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging Face Jobs cloud GPUs.

Covers COCO-format dataset preparation, Albumentations augmentation, mAP/mAR evaluation, accuracy metrics, SAM segmentation with bbox/point prompts, DiceCE loss, hardware selection, cost estimation, Trackio monitoring, and Hub persistence.

Use it when

Users mention training object detection, image classification, SAM, SAM2, segmentation, image matting, DETR, D-FINE, RT-DETR, ViT, timm, MobileNet, ResNet, bounding box models, or fine-tuning vision models on Hugging Face Jobs.

Inputs and results

What goes in, what comes back

  1. You give

    A short brief describing what you need

    Add a specialist method to the work.

  2. You get

    A finished result to review

    The method guides the work with your files and context.

  3. You check

    The sources and open questions

    Review the output against your brief.

Questions and answers

Questions about huggingface-vision-trainer

Updated May 13, 2026.

What does huggingface-vision-trainer do?

Trains and fine-tunes vision models for object detection (D-FINE, RT-DETR v2, DETR, YOLOS), image classification (timm models — MobileNetV3, MobileViT, ResNet, ViT/DINOv3 — plus any Transformers classifier), and SAM/SAM2 segmentation using Hugging Face Transformers on Hugging Face Jobs cloud GPUs. Covers COCO-format dataset preparation, Albumentations augmentation, mAP/mAR evaluation, accuracy metrics, SAM segmentation with bbox/point prompts, DiceCE loss, hardware selection, cost estimation, Trackio monitoring, and Hub persistence. Use when users mention training object detection, image classification, SAM, SAM2, segmentation, image matting, DETR, D-FINE, RT-DETR, ViT, timm, MobileNet, ResNet, bounding box models, or fine-tuning vision models on Hugging Face Jobs.

Who is huggingface-vision-trainer for?

It is designed for teams that want to hand off recurring work and review the result before it is used.

What do I need to use it?

Nothing extra is needed. Sign in to Relay to run it with your own files and context.

What does huggingface-vision-trainer produce?

A finished result to review. Results come back with their sources and any open questions for you to check before anything is used.

How much does huggingface-vision-trainer cost?

It is free to use. Paid model usage, if any, is shown before a run.

Can an AI agent use it?

Yes. Agents can read its public definition through the Relay API and run it with a person’s authorization. See the For Agents page for how to connect.

Details

Trust, access and technical details

Trust and access

Visibility
Public: anyone can view
Status
Published and maintained
Pricing
Free
Medium riskSecurity review passedBehavior not evaluated
Reviewed version
cmoq1dk2f02rf…
Reviewed
May 13, 2026
Static signals
12 signals
Source provenance
Commit pinned

Security review is a static package assessment, not a guarantee of safety or a behavioral qualification.

Use it from code or an agent

Read the public definition, then check cost and inputs with a preflight before a run.

Open definition

POST /api/v1/runs/preflight

How agents connect to Relay →

Tags and source

Repo
huggingface/skills
License
Apache-2.0

Have a task this would help with?

Describe it. Relay plans the work, picks the right playbook and shows the cost before anything runs.

Loading request box