Install my trainer model

[AI-Elmanawy]

Search before asking

• I have searched the HUB issues and found no similar feature requests.

Description

I already trained yolo v12s on my dataset but I don't know how to upload it on hub and use it in test my dataset on android platform

Use case

No response

Additional

No response

[UltralyticsAssistant]

👋 Hello @AI-Elmanawy, thank you for raising an issue about Ultralytics HUB 🚀! This is an automated response — an Ultralytics engineer will review and assist you soon.

In the meantime, please visit our HUB Docs to learn more:

• Quickstart. Start training and deploying YOLO models with HUB in seconds.
• Datasets: Preparing and Uploading. Learn how to prepare and upload your datasets to HUB in YOLO format.
• Projects: Creating and Managing. Group your models into projects for improved organization.
• Models: Training and Exporting. Train YOLOv5 and YOLOv8 models on your custom datasets and export them to various formats for deployment.
• Integrations. Explore different integration options for your trained models, such as TensorFlow, ONNX, OpenVINO, CoreML, and PaddlePaddle.
• Ultralytics HUB App. Learn about the Ultralytics App for iOS and Android, which allows you to run models directly on your mobile device.
  • iOS. Learn about YOLO CoreML models accelerated on Apple's Neural Engine on iPhones and iPads.
  • Android. Explore TFLite acceleration on mobile devices.
• Inference API. Understand how to use the Inference API for running your trained models in the cloud to generate predictions.

If this is a 🐛 Bug Report, please provide screenshots and steps to reproduce your problem to help us get started working on a fix.

If this is a ❓ Question (e.g., how to upload your trained model to HUB and use it on Android), please share:

• The YOLO model variant you trained and how you trained it (Ultralytics package version and command)
• The export format you intend to use (e.g., TFLite, ONNX, CoreML) and any export logs
• Whether you’ve tried uploading via HUB UI or API and what steps you took
• Your target Android setup (HUB App vs. custom Android app using exported model)
• Any screenshots or error messages you encountered

We try to respond to all issues as promptly as possible. Thank you for your patience! 🙌

[pderrenger]

Hi @AI-Elmanawy — great question!

Short answer

• Ultralytics HUB and the HUB App currently support YOLOv5, YOLOv8 and YOLO11 models on mobile. YOLO12 is not supported in HUB/App, and we recommend switching to YOLO11 for stability, memory efficiency and fast CPU/mobile inference.
• To test on Android you have two easy paths: use the Ultralytics HUB App for real‑time camera testing, or export your model to TFLite from HUB and run it inside your own Android app.

Recommended workflow (fastest path to Android)

1. Train or fine‑tune a YOLO11 model in HUB

• Go to Models → Train Model and select your dataset and a YOLO11 variant. You can train via Ultralytics Cloud, Google Colab, or Bring Your Own Agent. See the HUB Models guide for the end‑to‑end flow. See Train a model in the HUB Models docs.

2. Test on Android

• Option A: Install the Ultralytics HUB App from Google Play, sign in, pick your trained model, and run it live with your phone’s camera. See the HUB App overview.
• Option B: From your model’s Deploy tab in HUB, export to TFLite, then integrate with your Android project following the TFLite Integration Guide.

Helpful references

• Getting started and model training in HUB: See the Ultralytics HUB Quickstart and HUB Models docs.
• HUB App (Android/iOS) for real‑time testing: See the Ultralytics HUB App docs (supports YOLOv5, YOLOv8, and YOLO11).
• TFLite export and Android integration: See the TFLite Integration Guide.

About “uploading” an already‑trained local model

• HUB is designed around creating models in HUB (Cloud/Colab/BYOA). If you trained locally with the Ultralytics library, the easiest way to “bring it into HUB” is to start a short Bring Your Own Agent run from the Train Model dialog and point it at your dataset; the agent streams checkpoints and metrics to HUB automatically so your model appears in the UI. See the Train Model dialog → Bring your own agent in the HUB Models docs.
• There is also a low‑level SDK endpoint that uploads checkpoints to an existing HUB model (used by BYOA under the hood): upload_model(id, epoch, weights, is_best=False, map=0.0, final=False). See the SDK reference for upload_model. Note you must have a model ID created in HUB to use this.

Next steps / to unblock you

• Are you okay switching to YOLO11 for Android? If yes, I’ll outline the minimal steps for fine‑tuning on your dataset via HUB so you can test on your phone in minutes.
• If you must use your current YOLO12 weights, please share where they were trained (repo/framework) and the output format (e.g., .pt). We can advise if a compatible conversion path exists, but for HUB/App we strongly recommend YOLO11.

Thanks again for raising this — huge credit to the YOLO community and the Ultralytics team for the tooling that makes this smooth.

[AI-Elmanawy]

Thanks so much for your response but I have to used my trained yolo v12 model which already trained on PC using GPU with extend .pt Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: Paula Derrenger ***@***.***> Sent: Monday, November 3, 2025 8:24:55 PM To: ultralytics/hub ***@***.***> Cc: DR / Ahmed islam mahmoud helmy Elmanawy ***@***.***>; Mention ***@***.***> Subject: Re: [ultralytics/hub] Install my trainer model (Issue #1228) [https://avatars.githubusercontent.com/u/107626595?s=20&v=4]pderrenger left a comment (ultralytics/hub#1228)<#1228 (comment)> Hi @AI-Elmanawy<https://github.com/AI-Elmanawy> — great question! Short answer * Ultralytics HUB and the HUB App currently support YOLOv5, YOLOv8 and YOLO11 models on mobile. YOLO12 is not supported in HUB/App, and we recommend switching to YOLO11 for stability, memory efficiency and fast CPU/mobile inference. * To test on Android you have two easy paths: use the Ultralytics HUB App for real‑time camera testing, or export your model to TFLite from HUB and run it inside your own Android app. Recommended workflow (fastest path to Android) 1. Train or fine‑tune a YOLO11 model in HUB * Go to Models → Train Model and select your dataset and a YOLO11 variant. You can train via Ultralytics Cloud, Google Colab, or Bring Your Own Agent. See the HUB Models guide for the end‑to‑end flow. See Train a model in the HUB Models docs. 1. Test on Android * Option A: Install the Ultralytics HUB App from Google Play, sign in, pick your trained model, and run it live with your phone’s camera. See the HUB App overview. * Option B: From your model’s Deploy tab in HUB, export to TFLite, then integrate with your Android project following the TFLite Integration Guide. Helpful references * Getting started and model training in HUB: See the Ultralytics HUB Quickstart and HUB Models docs. * HUB App (Android/iOS) for real‑time testing: See the Ultralytics HUB App docs (supports YOLOv5, YOLOv8, and YOLO11). * TFLite export and Android integration: See the TFLite Integration Guide. About “uploading” an already‑trained local model * HUB is designed around creating models in HUB (Cloud/Colab/BYOA). If you trained locally with the Ultralytics library, the easiest way to “bring it into HUB” is to start a short Bring Your Own Agent run from the Train Model dialog and point it at your dataset; the agent streams checkpoints and metrics to HUB automatically so your model appears in the UI. See the Train Model dialog → Bring your own agent in the HUB Models docs. * There is also a low‑level SDK endpoint that uploads checkpoints to an existing HUB model (used by BYOA under the hood): upload_model(id, epoch, weights, is_best=False, map=0.0, final=False). See the SDK reference for upload_model. Note you must have a model ID created in HUB to use this. Next steps / to unblock you * Are you okay switching to YOLO11 for Android? If yes, I’ll outline the minimal steps for fine‑tuning on your dataset via HUB so you can test on your phone in minutes. * If you must use your current YOLO12 weights, please share where they were trained (repo/framework) and the output format (e.g., .pt). We can advise if a compatible conversion path exists, but for HUB/App we strongly recommend YOLO11. Thanks again for raising this — huge credit to the YOLO community and the Ultralytics team for the tooling that makes this smooth. — Reply to this email directly, view it on GitHub<#1228 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AUBQH5JYNFINM2URTV3T7PT325CRPAVCNFSM6AAAAACK6DU6ICVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTIOBQGI3DINJWHA>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[pderrenger]

Thanks for confirming, Ahmed — totally get the constraint.

Current status (as of Nov 4, 2025)

• Ultralytics HUB and the HUB App support YOLOv5, YOLOv8 and YOLO11 for training, preview and mobile testing. YOLO12 models aren’t supported for HUB preview/export or the Android/iOS apps, so a YOLO12 .pt can’t be tested on Android via HUB today. See the Ultralytics HUB overview and supported models list in the HUB docs and the HUB App page for details:
  • Ultralytics HUB overview (supported models): See the HUB landing page in docs.
  • HUB App (Android/iOS) supported models: See the Ultralytics HUB App docs.

Practical options

1. Fastest path to Android via HUB (recommended)

• Fine‑tune a YOLO11 variant on your dataset in HUB (Cloud, Colab or BYOA). Then:
  • Preview on your phone with the HUB App.
  • Or export to TFLite from the model’s Deploy tab for your own Android app.
• References:
  • Getting started with models/training in HUB: See the HUB Models guide.
  • Export to TFLite for Android: See the YOLO11 to TFLite export guide.

2. If you must keep your exact YOLO12 .pt

• Storage in HUB (artifact only): You can upload the checkpoint to a HUB model with the HUB‑SDK so you have it versioned and shareable, but it will not run/preview/export in HUB.
  • Minimal example:
    • pip install hub-sdk
    • from hub_sdk import HUBClient
    • client = HUBClient({"api_key": ""})
    • model = client.model("<MODEL_ID>") # create in UI or via SDK first
    • model.upload_model(epoch=0, weights="path/to/your_yolo12.pt", is_best=True, final=True)
  • See the HUB‑SDK model management docs (upload_model) for details.
• Android outside HUB: If you plan to integrate YOLO12 directly in your own app (e.g., with a runtime you manage), that’s outside HUB/App support. We can’t provide one‑to‑one mobile integration support here, but keeping the checkpoint in HUB (above) still helps with artifact management and collaboration.

Why we suggest YOLO11 for mobile

• YOLO12 isn’t recommended for production; YOLO11 is the latest stable Ultralytics YOLO with better training stability, lower memory use, and faster CPU/mobile inference. You’ll also get one‑click HUB exports (TFLite, ONNX, CoreML) and the HUB App preview.

If you’re open to migrating, I’m happy to outline a minimal YOLO11 fine‑tune plan (epochs/image size/variant) to match your current accuracy, then show where to tap Export→TFLite for Android. If you prefer to keep the YOLO12 .pt, I can help you set up the HUB‑SDK “storage‑only” flow so the file lives under a model in HUB. 🙏

Credits to the YOLO community and the Ultralytics team for the tooling that makes these paths smooth.

[glenn-jocher]

Hi @AI-Elmanawy, following up on this since HUB winds down on July 31, 2026: Ultralytics Platform now covers this workflow — you can drag-and-drop trained .pt checkpoints (up to 1 GB) into a project and export to 19+ formats including LiteRT and NCNN for Android with a single click (model docs). The documented architectures are YOLOv5/v8/11/26, so a YOLO11 or YOLO26 retrain of your dataset would be supported end-to-end there.

For the Android side, there are open-source demo apps with full source code you can drop your exported model into: ultralytics/yolo-flutter-app (Android + iOS) and ultralytics/yolo-ios-app (native iOS).