Question 1

What is TjMakeBot?

Accepted Answer

TjMakeBot is a free AI-powered image annotation tool designed for creating training datasets for object detection models like YOLO, Faster R-CNN, and SSD. It features a built-in AI Assistant that allows you to annotate images automatically through natural language chat, making the annotation process faster and more efficient.

Question 2

How do I use the AI Assistant for automatic annotation?

Accepted Answer

Click the "AI Assistant" button in the toolbar, then describe what you want to annotate in natural language. For example, you can say "annotate all people" or "label all cars and trucks". The AI will automatically detect and annotate the objects. You can also use "Apply to All Images" to batch process multiple images at once.

Question 3

What annotation formats does TjMakeBot support?

Accepted Answer

TjMakeBot supports multiple annotation formats including YOLO (txt with normalized coordinates), Pascal VOC (XML with absolute pixel coordinates), COCO JSON, and CSV. You can import existing annotations and export in multiple formats simultaneously. The tool supports YOLOv5, YOLOv8, YOLOv9, YOLOv10, and YOLOv11 formats.

Question 4

支援哪些3D點雲格式？

Accepted Answer

TjMakeBot在3D標註編輯器中支援PCD和PLY等常見點雲格式。

Question 5

How do I draw and adjust 3D bounding boxes?

Accepted Answer

In 3D mode, press B to enter draw mode and click three times to create a 3D box. Then select the object and use W/E/R to switch translate, rotate, and scale. You can also fine-tune center, dimensions, and rotation values in the properties panel.

Question 6

可以匯出3D標註用於模型訓練嗎？

Accepted Answer

是的。3D標註可以匯出為KITTI相容和JSON格式，適用於大多數3D檢測訓練流程。

Question 7

How do I export annotations in YOLO format?

Accepted Answer

After annotating your images, click the "Export Annotations" button. Select the YOLO format option, and the tool will generate a ZIP file containing all annotation files (.txt) and images organized in the standard YOLO dataset structure (images/ and labels/ folders).

Question 8

Is TjMakeBot free to use?

Accepted Answer

Yes, TjMakeBot offers free basic features including manual annotation, import/export in multiple formats, video decoding, and batch processing. The AI Assistant feature requires registration and uses a token-based system. New users receive free AI credits to get started.

Question 9

What languages does TjMakeBot support?

Accepted Answer

TjMakeBot supports multiple interface languages including English, Chinese (Simplified and Traditional), Japanese, Korean, German, French, and Russian. You can switch languages using the language selector in the top navigation bar.

Question 10

How do I batch process multiple images?

Accepted Answer

You can upload multiple images at once, and use the AI Assistant's "Apply to All Images" feature to automatically annotate all images with the same instruction. You can also use the batch detection feature with trained models to automatically detect objects across all images.

Question 11

How does the video-to-frames feature work?

Accepted Answer

TjMakeBot includes a video decoder that can extract frames from videos at your specified FPS (frames per second). Upload a video file, set your desired FPS (e.g., 1 FPS for static scenes, 5-10 FPS for motion), and the tool will automatically decode the video into individual image frames that you can then annotate.

Question 12

How are annotations saved?

Accepted Answer

For logged-in users, annotation projects are automatically saved to the cloud every 30 seconds. All images, labels, and annotations are included. You can access your projects from any device. The system also supports offline editing with automatic sync. Guest users can export annotations manually in YOLO/VOC/COCO/CSV formats.

Question 13

How can I contact technical support?

Accepted Answer

You can contact our technical support team by email at service@taojsh.com. We typically respond within 1-2 business days. You can also check the Tutorials page for detailed guides and tips.

Question 14

TjMakeBot 支援團隊 review 和交付 handoff 嗎？

Accepted Answer

支援。TjMakeBot 已支援共享專案、review 隊列、issue 處理、dataset version 凍結、release 交付、客戶分享頁，以及面向團隊協作生產的審計型 handoff 流程。

Question 15

團隊可以管理 dataset version 和訓練 lineage 嗎？

Accepted Answer

可以。approved output 可以凍結成 dataset version，並和 release、training 流程關聯起來，讓團隊清楚說明版本來源、交付範圍和模型來源，而不是只靠臨時匯出文件。

Question 16

Team Review Cloud 和 Enterprise 應該怎麼選？

Accepted Answer

當你需要託管雲上的多人 review、交付流轉和可審計協作時，選擇 Team Review Cloud；當企业部署、SSO、採購流程、合約評審或更深的系統整合邊界變成硬需求時，就應該走 Enterprise 路徑。

Question 17

如果我們需要 API、webhook 或系統整合怎麼辦？

Accepted Answer

建議走企業詢價路徑，先確認目前 API、webhook、callback 以及外部系統整合的適配範圍，再對齊上線節奏和採購要求。

Question 18

支援企业部署嗎？

Accepted Answer

支援按企業場景評估企业部署，尤其適合那些需要安全評審、權限邊界、審計支持、採購協同或比託管雲更強部署控制的團隊。

Question 19

TjMakeBot 能支援 SSO 和面向審計的控制嗎？

Accepted Answer

當團隊對 SSO、租戶邊界、訪問治理、審計要求等能力有明確採購條件時，正確路徑是走企業評估，並結合 rollout 要求一起確認。

Question 20

What types of annotations does TjMakeBot support?

Accepted Answer

TjMakeBot supports multiple annotation types including bounding boxes (rectangles), polygons, oriented bounding boxes (OBB) for rotated objects, and keypoint annotations for pose estimation. You can switch between annotation types using the toolbar.

Question 21

What use cases is TjMakeBot suitable for?

Accepted Answer

TjMakeBot is ideal for autonomous driving dataset creation, industrial defect detection, medical image annotation, retail product recognition, security surveillance analysis, agricultural monitoring, and any computer vision research requiring labeled training data.

Question 22

What is cloud model training?

Accepted Answer

TjMakeBot provides cloud YOLO model training service. After completing data annotation, you can submit training tasks directly in the browser and use cloud GPU to train models without configuring local Python environment or purchasing GPU. Supports YOLOv5, YOLOv8, YOLOv9, YOLOv10, YOLOv11 and more.

Question 23

How is cloud training charged?

Accepted Answer

Cloud training is charged by training duration, deducted from your account balance. New users receive free training credits upon registration. The cost depends on model size, number of epochs, and dataset size. You can view estimated cost before submitting a task.

Question 24

How long does training take?

Accepted Answer

Training time depends on several factors: dataset size, model version, number of epochs, and image size. Generally, a small dataset (100-500 images) using a lightweight model (e.g., YOLOv8n) for 100 epochs typically takes 10-30 minutes. Larger datasets or complex models may take several hours.

Question 25

What is the training queue?

Accepted Answer

The training queue manages all user training tasks. When you submit a training task, it enters the queue waiting for GPU resources. You can view task status, estimated wait time, and queue position in the queue panel. Tasks are executed in submission order.

Question 26

What formats can I export after training?

Accepted Answer

After training completes, you can export the model in various formats: PyTorch (.pt) for research and continued training, ONNX for cross-platform inference, TensorRT for NVIDIA GPU acceleration, CoreML for iOS/macOS deployment, TFLite for Android deployment, OpenVINO for Intel hardware acceleration, NCNN for mobile and embedded devices.

Question 27

What if training fails?

Accepted Answer

If training fails, you can view error messages and training logs in task details. Common causes include: dataset format errors, insufficient annotation data, class imbalance, etc. After fixing the issue, click "Resubmit" to create a new training task. If the problem persists, contact technical support.

Question 28

Can I cancel an ongoing training?

Accepted Answer

Yes, you can cancel queued or training tasks anytime. Click the "Stop Training" button in the training queue or task details page. Cancelled tasks will not incur additional charges, but consumed training time will not be refunded.

Question 29

How does auto cloud storage work?

Accepted Answer

Once you start annotating, your project is automatically created and saved to the cloud every 30 seconds. All images, labels, and annotations are included. You can access your projects from any device by logging in. The system also supports offline editing — your work is saved locally and synced automatically when the connection is restored.

Question 30

How do I access my saved projects?

Accepted Answer

Click the folder icon in the toolbar to open the project list. You can view all your saved projects, open any project to continue working, rename projects by clicking the project name, or delete projects you no longer need.

Frequently Asked Questions

Enterprise, integration, or enterprise deployment questions?

Related Resources

Frequently Asked Questions

Enterprise, integration, or enterprise deployment questions?

What is TjMakeBot?+

How do I use the AI Assistant for automatic annotation?+

What annotation formats does TjMakeBot support?+

支援哪些3D點雲格式？+

How do I draw and adjust 3D bounding boxes?+

可以匯出3D標註用於模型訓練嗎？+

How do I export annotations in YOLO format?+

Is TjMakeBot free to use?+

What languages does TjMakeBot support?+

How do I batch process multiple images?+

How does the video-to-frames feature work?+

How are annotations saved?+

How can I contact technical support?+

TjMakeBot 支援團隊 review 和交付 handoff 嗎？+

團隊可以管理 dataset version 和訓練 lineage 嗎？+

Team Review Cloud 和 Enterprise 應該怎麼選？+

如果我們需要 API、webhook 或系統整合怎麼辦？+

支援企业部署嗎？+

TjMakeBot 能支援 SSO 和面向審計的控制嗎？+

What types of annotations does TjMakeBot support?+

What use cases is TjMakeBot suitable for?+

What is cloud model training?+

How is cloud training charged?+

How long does training take?+

What is the training queue?+

What formats can I export after training?+

What if training fails?+

Can I cancel an ongoing training?+

How does auto cloud storage work?+

How do I access my saved projects?+

Related Resources