-
+
## Before Start
@@ -82,17 +140,6 @@ Purchase Links:[Sipeed Taobao](https://item.taobao.com/item.htm?id=846226367137)
>! Note that currently only the MaixCAM development board is supported. Other development boards with the same chip are not supported, including Sipeed's development boards with the same chip. Please be careful not to purchase the wrong board, which could result in unnecessary waste of time and money.
-
-## Getting Started
-
-Please select the documentation corresponding to your hardware platform to proceed:
-
-| Hardware Platform | Getting Started Guide |
-|-|-|
-|MaixCAM Lite|[Quick Start MaixCAM(Screenless Version)](./README_no_screen.md)|
-|MaixCAM/MaixCAM Pro|[Quick Start MaixCAM](./README_MaixCAM.md)|
-|MaixCAM2|[Quick Start MaixCAM2](./README_MaixCAM2.md)|
-
## Next Steps
If you like what you've seen so far, **please be sure to give the MaixPy open-source project a star on [GitHub](https://github.com/sipeed/MaixPy) (you need to log in to GitHub first). Your star and recognition is the motivation for us to continue maintaining and adding new features!**
diff --git a/docs/doc/en/README_MaixCAM.md b/docs/doc/en/README_MaixCAM.md
index d25aaa0c..f9e09867 100644
--- a/docs/doc/en/README_MaixCAM.md
+++ b/docs/doc/en/README_MaixCAM.md
@@ -7,6 +7,12 @@ title: MaixCAM MaixPy Quick Start
width: 100%;
display: table;
}
+ .biliiframe {
+ width: 100%;
+ min-height: 30em;
+ border-radius: 0.5em;
+ border: 1em solid white;
+ }
@media screen and (max-width: 900px){
#head_links th, #head_links td {
@@ -19,6 +25,10 @@ title: MaixCAM MaixPy Quick Start
## Getting Started
+### Getting Started Video
+
+
+
### Prepare the TF Image Card and Insert it into the Device
If the package you purchased includes a TF card, it already contains the factory image. If the TF card was not installed in the device at the factory, you will first need to carefully open the case (be careful not to tear the ribbon cables inside) and then insert the TF card. Additionally, since the firmware from the factory may be outdated, it is highly recommended to follow the instructions on [Upgrading and Flashing the System](./basic/os.md) to upgrade the system to the latest version.
@@ -158,4 +168,4 @@ Here is the translation:
If you want the program to start automatically on boot, you can set it in `Settings -> Boot Startup`.
- More MaixVision usage refer to [MaixVision documentation](./basic/maixvision.md)。
\ No newline at end of file
+ More MaixVision usage refer to [MaixVision documentation](./basic/maixvision.md)。
diff --git a/docs/doc/en/README_MaixCAM2.md b/docs/doc/en/README_MaixCAM2.md
index 0295eae2..ec0314fa 100644
--- a/docs/doc/en/README_MaixCAM2.md
+++ b/docs/doc/en/README_MaixCAM2.md
@@ -7,6 +7,12 @@ title: MaixCAM2 MaixPy Quick Start
width: 100%;
display: table;
}
+ .biliiframe {
+ width: 100%;
+ min-height: 30em;
+ border-radius: 0.5em;
+ border: 1em solid white;
+ }
@media screen and (max-width: 900px){
#head_links th, #head_links td {
@@ -19,7 +25,11 @@ title: MaixCAM2 MaixPy Quick Start
## Getting Started
->! MaixCAM2 has built-in eMMC storage, so it can operate without requiring a TF card. If you need to upgrade or flash the system, please refer directly to the System Upgrade and Flashing guide.
+>! MaixCAM2 is available in versions with and without eMMC storage. The 32GB eMMC version normally boots from eMMC and does not require a TF card for daily use; the version without eMMC must use a TF card with a flashed system to boot. A TF card can also be used to boot from TF card, or to flash or recover the system to eMMC. For system upgrade or flashing, see System Upgrade and Flashing.
+
+### Getting Started Video
+
+
### Power On
@@ -154,4 +164,4 @@ Here is the translation:
If you want the program to start automatically on boot, you can set it in `Settings -> Boot Startup`.
- More MaixVision usage refer to [MaixVision documentation](./basic/maixvision.md)。
\ No newline at end of file
+ More MaixVision usage refer to [MaixVision documentation](./basic/maixvision.md)。
diff --git a/docs/doc/en/ai_model_converter/ai_model_deploy.md b/docs/doc/en/ai_model_converter/ai_model_deploy.md
new file mode 100644
index 00000000..2479920e
--- /dev/null
+++ b/docs/doc/en/ai_model_converter/ai_model_deploy.md
@@ -0,0 +1,15 @@
+---
+title: AI Model Download, Debugging, and Deployment Guide
+---
+
+## Choose a Model Deployment Workflow
+
+Before deploying a local model on MaixCAM2, first identify the model source and deployment path. Choose the workflow that matches your current resources instead of starting with ONNX conversion immediately.
+
+| Goal | Recommended workflow | Documentation |
+| --- | --- | --- |
+| Use built-in or ready-made models | Use built-in models first. For more resolutions or class sets, filter MaixCAM2 models in [MaixHub Model Zoo](https://maixhub.com/model/zoo?platform=maixcam2). Model packages usually include a `.mud` file and corresponding `.axmodel` files, which should be placed in the same directory on the device | [Model and dataset sources](../pro/datasets.md) |
+| Train a custom recognition target | Use MaixHub online training to complete data collection, annotation, training, and deployment | [MaixHub online training](../vision/maixhub_train.md) |
+| Deploy an ONNX model | Follow the MaixCAM2 conversion workflow to convert ONNX into `.mud` + `.axmodel` files before deployment | [MaixCAM2 model conversion](./maixcam2.md) |
+
+After choosing your workflow, continue with the corresponding document.
diff --git a/docs/doc/en/ai_model_converter/maixcam2.md b/docs/doc/en/ai_model_converter/maixcam2.md
index 81263de7..66cb7f6b 100644
--- a/docs/doc/en/ai_model_converter/maixcam2.md
+++ b/docs/doc/en/ai_model_converter/maixcam2.md
@@ -10,6 +10,8 @@ Models trained on a computer cannot be directly used on MaixCAM2 due to its limi
This article describes how to convert an ONNX model into a model usable by MaixCAM2 (MUD model).
+If you are not sure whether to download an existing model, train online, or convert ONNX yourself, read the [beginner model download, debugging, and deployment guide](./ai_model_deploy.md) first.
+
## Supported Model File Format for MaixCAM2
MUD (Model Universal Description) is a model description file supported by MaixPy that unifies models across different platforms, making MaixPy code cross-platform. It is a text file in `ini` format and can be edited with a text editor.
@@ -178,10 +180,50 @@ Edit as per your model. For YOLO11:
The `[basic]` section is required. Once set, you can load and run the model using `maix.nn.NN` in `MaixPy` or `MaixCDK`.
+## Deploy to the Device and Verify Quickly
+
+MaixPy usually loads the `.mud` file. The `.mud` file then points to the actual `.axmodel` files. The simplest approach is to put them in the same directory, for example:
+
+```text
+/root/models/my_model.mud
+/root/models/my_model_npu.axmodel
+/root/models/my_model_vnpu.axmodel
+```
+
+Then first confirm that the model can be loaded:
+
+```python
+from maix import nn
+
+model = nn.NN("/root/models/my_model.mud")
+print(model)
+```
+
+If the model type is already supported by MaixPy, prefer the wrapped API. For example, for YOLO, see the [YOLO object detection documentation](../vision/yolov5.md):
+
+```python
+from maix import nn
+
+detector = nn.YOLO11(model="/root/models/yolo11n.mud", dual_buff=True)
+```
+
+## Debug Checklist
+
+If the model cannot be loaded or the result is wrong, check these items first:
+
+1. Whether the `.mud` and `.axmodel` files are in the same directory, and whether `model_npu` and `model_vnpu` in the `.mud` file use the correct file names.
+2. Whether the path used on the device really exists, such as `/root/models/xxx.mud`.
+3. Whether `labels` exactly matches the class count and class order used during training.
+4. Whether `model_type` is supported by MaixPy, such as `yolo11`, `yolov8`, or `classifier`.
+5. Whether input resolution, `mean`, `scale`, and RGB/BGR order match training, export, and conversion settings.
+6. Whether the ONNX output nodes exactly match `output_processors` in `config.json`.
+7. If you are not sure whether the problem is the model or your code, test with a MaixHub model or a built-in model first. After an official model runs correctly, debug your own model.
+
+After these basic checks pass, continue with the specific model documentation or conversion workflow.
+
If MaixPy doesn't support your model, define your own `extra` fields and write decoding logic. You can either:
* Use Python in `MaixPy` to load the model via `maix.nn.NN`, run `forward`/`forward_image`, and process the outputs in Python (easier but slower);
* Or, for better performance and reusable integration, write C++ logic in `MaixCDK`, see [YOLOv5 example](https://github.com/sipeed/MaixCDK/blob/71d5b3980788e6b35514434bd84cd6eeee80d085/components/nn/include/maix_nn_yolov5.hpp).
Once done, consider submitting a PR to `MaixPy` or share your model on [MaixHub](https://maixhub.com/share) to earn rewards ranging from ¥30 to ¥2000!
-
diff --git a/docs/doc/en/basic/upgrade.md b/docs/doc/en/basic/upgrade.md
index 1714c26c..06444737 100644
--- a/docs/doc/en/basic/upgrade.md
+++ b/docs/doc/en/basic/upgrade.md
@@ -8,20 +8,24 @@ First, let's distinguish between **`System`** and **`MaixPy`**:
* **System**: The foundation for running all software, including the operating system and drivers, serving as the cornerstone for software operation.
* **MaixPy**: A software package that relies on system drivers to function.
-## Getting the Latest System
+## Get the Latest System and Flash It to Hardware
-Find the latest system image files on the [MaixPy Releases page](https://github.com/sipeed/MaixPy/releases), for example:
+Before flashing, confirm both the device model and where the system is stored. MaixCAM / MaixCAM-Pro run the system from a TF card, so a TF card is required. MaixCAM2 normally runs from onboard eMMC, so a TF card is usually not required for regular updates. For regular updates, use the USB flashing workflow in the corresponding flashing page first; when booting from a TF card, recovering the system, or replacing storage media, follow the method described on that page.
-* `maixcam_os_20240401_maixpy_v4.1.0.xz`: MaixCAM system image including MaixPy v4.1.0.
-* `maixcam-pro_os_20240401_maixpy_v4.1.0.xz`: MaixCAM Pro system image including MaixPy v4.1.0.
-* `maixcam2_os_20250801_maixpy_v4.11.0.xz`: MaixCAM2 system image including MaixPy v4.11.0. Since the MaixCAM system image exceeds 2GB, it will only be available on Sourceforge. Users in China can download it via the shared files in the QQ group.
+After confirmation, use the table below to choose the correct system image and flashing page.
-**Make sure to download the system image that corresponds to your device model**. Downloading the wrong image may cause device damage.
+| Device | Download entry | File to choose | Flashing page |
+| --- | --- | --- | --- |
+| MaixCAM | [MaixPy Releases page](https://github.com/sipeed/MaixPy/releases) | `maixcam-*.img.xz` | [MaixCAM System Flashing](https://wiki.sipeed.com/hardware/zh/maixcam/os.html) |
+| MaixCAM-Pro | [MaixPy Releases page](https://github.com/sipeed/MaixPy/releases) | `maixcam-pro-*.img.xz` | [MaixCAM System Flashing](https://wiki.sipeed.com/hardware/zh/maixcam/os.html) |
+| MaixCAM2 | [Baidu Netdisk (extraction code: vjex)](https://pan.baidu.com/s/1r4ECNlaTVxhWIafNBZOztg) | `maixcam2-*-maixpy-*_sd.img.7z.*` | [MaixCAM2 System Flashing](https://wiki.sipeed.com/hardware/zh/maixcam/maixcam2_os.html) |
-> Users in China with slow download speeds can use tools like Xunlei for faster downloads.
-> Alternatively, use proxy sites such as [github.abskoop.workers.dev](https://github.abskoop.workers.dev/) for downloads.
+**Make sure to download the system image that corresponds to your device model**. Downloading the wrong image may cause abnormal behavior and may require recovery flashing.
+
+When downloading, choose the newest file that matches the "File to choose" column in the table. For MaixCAM and MaixCAM-Pro, select the corresponding image on the releases page. MaixCAM2 images are split 7z archives; download all `.7z.00x` parts of the same version from Baidu Netdisk before extracting and flashing.
-Backup mirror: [Sourceforge](https://sourceforge.net/projects/maixpy/files/) (may not be up-to-date, so prefer the above official sources)
+
+> Users in China with slow download speeds can use tools like Xunlei for faster downloads.
## Backup Your Data
@@ -37,18 +41,6 @@ Backup methods:
* Remove the storage media and use a card reader to copy files directly. Note: the root filesystem is formatted as `ext4`, which Windows does not support by default (you can use third-party software like DiskGenius to read it).
-## Flashing the System to Hardware
-
-| Item | MaixCAM / MaixCAM-Pro | MaixCAM2 |
-| --------------- | ---------- | ------ |
-| Flashing Docs | [MaixCAM System Flashing](https://wiki.sipeed.com/hardware/zh/maixcam/os.html) | [MaixCAM2 System Flashing](https://wiki.sipeed.com/hardware/zh/maixcam/maixcam2_os.html) |
-| System Storage | TF Card | Built-in EMMC (/TF Card) |
-| TF Card Required | Yes | No |
-| Flashing Method | USB flashing or card reader flashing | USB flashing or card reader flashing |
-| Recommended Method | USB flashing | USB flashing |
-| Recovery Flashing | Card reader flashing | USB flashing / card reader flashing |
-
-
## When to Update the System vs. Updating MaixPy Only
To simplify the process and avoid issues, it is **recommended to update the system whenever upgrading MaixPy**.
diff --git a/docs/doc/en/pro/datasets.md b/docs/doc/en/pro/datasets.md
index 8e58ed23..7388ea35 100644
--- a/docs/doc/en/pro/datasets.md
+++ b/docs/doc/en/pro/datasets.md
@@ -8,6 +8,14 @@ title: MaixCAM MaixP Where to Find Models and Datasets
Visit the [MaixHub Model Library](https://maixhub.com/model/zoo) and filter by the corresponding hardware platform to find suitable models.
+If you only want to verify that the workflow works, start with a built-in model. The system usually includes common models in `/root/models`, so you can reference the corresponding `.mud` file directly in your code without downloading anything.
+
+For more models, open [MaixHub Model Zoo](https://maixhub.com/model/zoo?platform=maixcam2), filter by the MaixCAM2 platform, and download a model package. After downloading, check whether it contains:
+
+- `.mud`: the model description file. MaixPy code usually loads this file.
+- `.axmodel`: the actual model file that runs on MaixCAM2.
+- Example code or instructions: if the model page provides an example, run that first.
+
## What Are Datasets Used For?
First, check the [MaixHub Model Library](https://maixhub.com/model/zoo) to see if there’s a model you need. If not, you can train your own model. Training a model requires a dataset, which provides the data needed for training.
diff --git a/docs/doc/en/sidebar.yaml b/docs/doc/en/sidebar.yaml
index 2d9d4439..e10cd422 100644
--- a/docs/doc/en/sidebar.yaml
+++ b/docs/doc/en/sidebar.yaml
@@ -137,24 +137,26 @@ items:
label: Whisper Speech-Recognition Model
- file: mllm/tts_melotts.md
label: MeloTTS Speech-Synthesis Model
-- label: Local NPU deployment of AI models
+- label: Local deployment of AI models
items:
+ - file: ai_model_converter/ai_model_deploy.md
+ label: AI model download and deployment guide
+ - file: pro/datasets.md
+ label: Where to find models & datasets
+ - file: vision/maixhub_train.md
+ label: MaixHub online AI training
+ - file: vision/customize_model_yolo.md
+ label: YOLO model offline training
- file: ai_model_converter/maixcam2.md
- label: ONNX model to MaixCAM2's
+ label: Convert ONNX model for MaixCAM2
- file: ai_model_converter/maixcam.md
- label: ONNX model to MaixCAM's
+ label: Convert ONNX model for MaixCAM
- file: ai_model_converter/onnx_export.md
label: Trim ONNX model output nodes
- file: ai_model_converter/web_converter.md
label: Convert YOLO models with Web UI
- - file: pro/datasets.md
- label: Where to find models & datasets
- file: pro/customize_model.md
- label: Customize new AI model
- - file: vision/maixhub_train.md
- label: MaixHub online AI training
- - file: vision/customize_model_yolo.md
- label: YOLO model offline training
+ label: Port a new AI model
diff --git a/docs/doc/en/vision/maixhub_train.md b/docs/doc/en/vision/maixhub_train.md
index 741d8542..ddd8d068 100644
--- a/docs/doc/en/vision/maixhub_train.md
+++ b/docs/doc/en/vision/maixhub_train.md
@@ -1,52 +1,136 @@
---
-title: Using MaixHub to Train AI Models for MaixCAM MaixPy
-update:
- - date: 2024-04-03
- author: neucrack
- version: 1.0.0
- content: Initial document
+title: Train AI Models Online with MaixHub
---
## Introduction
-MaixHub offers the functionality to train AI models online, directly within a browser. This eliminates the need for expensive hardware, complex development environments, or coding skills, making it highly suitable for beginners as well as experts who prefer not to delve into code.
+[MaixHub](https://maixhub.com/) provides online AI model training. You can collect data, upload images, annotate images, train models, and deploy models in a browser without setting up a local training environment or configuring a GPU.
-## Basic Steps to Train a Model Using MaixHub
+This page follows the official example video workflow and uses an **image detection model** as the example. If you only want to try AI features quickly, first check the [MaixHub Model Zoo](https://maixhub.com/model/zoo) to see whether a ready-to-use model already exists. Use online training when you need to recognize custom targets.
-### Identify the Data and Model Types
+
-To train an AI model, you first need to determine the type of data and model. As of April 2024, MaixHub provides models for image data including `Object Classification Models` and `Object Detection Models`. Object classification models are simpler than object detection models, as the latter require marking the position of objects within images, which can be more cumbersome. Object classification merely requires identifying what is in the image without needing coordinates, making it simpler and recommended for beginners.
+> The screenshots below are for workflow guidance. Account, project, dataset, image file names, QR codes, and training task IDs have been blurred.
-### Collect Data
+## Official Video Demos
-As discussed in AI basics, training a model requires a dataset for the AI to learn from. For image training, you need to create a dataset and upload images to it.
+MaixHub provides two official videos on the home page. Watch **Quick Start** first to understand the overall online training workflow. When you need to follow the page step by step, watch **Tutorial** as a detailed walkthrough. After logging in to [MaixHub](https://maixhub.com/), click **Play Video** in the video area at the top of the home page to watch.
-Ensure the device is connected to the internet (WiFi).
-Open the MaixHub app on your device and choose to collect data to take photos and upload them directly to MaixHub. You need to create a dataset on MaixHub first, then click on device upload data, which will display a QR code. Scan this QR code with your device to connect to MaixHub.
+
-It's important to distinguish between training and validation datasets. To ensure the performance during actual operation matches the training results, the validation dataset must be of the same image quality as those taken during actual operation. It's also advisable to use images taken by the device for the training set. If using internet images, restrict them to the training set only, as the closer the dataset is to actual operational conditions, the better.
+## MaixHub Model Training Workflow
+
+This section keeps the complete tutorial in one workflow category and follows the actual operation order: create a project, prepare data, annotate images, train the model, and deploy it.
+
+| Step | Operation |
+| --- | --- |
+| Create a project | Select the model type and target hardware platform |
+| Prepare a dataset | Upload training and validation images from the device or local files |
+| Annotate images | Detection models require boxes for each target |
+| Create a training task | Select the model, augmentation, and training parameters |
+| Check training results | Review curves and validation examples |
+| Deploy to device | Download the model package and upload it to the device with MaixVision |
+
+### Create a Training Project
+
+From the MaixHub home page, enter **Model Training** and create a new training project. Select the model type and hardware platform during project creation:
+
+
+
+This guide uses an **image detection model** in the following steps. It is suitable when the model needs to locate target positions in an image.
+
+After the project is created, follow the left navigation to complete dataset preparation, annotation, training, and deployment.
+
+### Prepare the Dataset
+
+Create a dataset after entering the project. The data type and annotation type must match the project. For example, a detection project should use image data and detection annotation.
+
+It is recommended to collect images with the MaixHub app on the device. Device-captured images are closer to the real camera angle, resolution, and lighting used during deployment, so the trained model is more likely to run reliably on the device.
+
+On the web page, open **Collect Data**, choose whether the images should go to the training set or validation set, and generate a QR code:
+
+
+
+Basic workflow:
+
+1. Make sure the device is connected to WiFi.
+2. Create and select a dataset on the web page.
+3. Choose the training set or validation set, then generate the QR code.
+4. Open the MaixHub app on the device, scan the QR code, and upload captured images.
+
+The training set is used for learning target features, while the validation set is used for evaluating training quality. For detection models, keep at least 5 validation images for each label, otherwise training may not start. The validation set should not duplicate the training set and should use real-scene images whenever possible.
+
+In the dataset page, select images in batches and move them to the training set or validation set as needed:
+
+
### Annotate Data
-For classification models, images are annotated during upload by selecting the appropriate category for each image.
+Image classification only requires selecting a category for each image. Image detection requires drawing boxes for targets and assigning labels to those boxes.
+
+Open **Annotate Data**, create annotations, draw target boxes, and save the result:
+
+
+
+Annotation tips:
+
+* Keep each box close to the target object and avoid including too much background.
+* Use consistent annotation rules for the same target class.
+* Do not miss targets that should be detected.
+* Leave blurry, heavily occluded, or uncertain images out of the training set first.
+
+For larger datasets, complete one full training and deployment cycle with a small dataset first, then gradually add more data to improve results.
+
+### Create a Training Task
+
+After checking the dataset and annotations, open **Create Task**. The page mainly contains image augmentation, model selection, and training parameter panels:
+
+
+
+Suggestions for beginners:
+
+* Select the actual deployment device, such as MaixCAM or MaixCAM2.
+* Use the default recommended model network first.
+* Enable image augmentation as needed, but do not use it as a replacement for real-scene data.
+* Enable data balancing when the number of images differs greatly between classes.
+* Add negative samples when background false positives need to be reduced.
+
+After confirming the model information and parameters, click **Create Training Task**, enter a task name, and start training.
+
+### Check Training Results
+
+After training starts, open **Training Records** to view progress, logs, dataset statistics, and training parameters. When training is complete, the result page shows loss curves, accuracy curves, and validation examples:
+
+
+
+Check the following first:
+
+* Whether the curves are stable and free of obvious anomalies.
+* Whether the validation examples are recognized correctly.
+* Whether wrong results are concentrated under certain angles, lighting conditions, or backgrounds.
+
+If training fails, check the log on the right first. In the example below, the failure is caused by fewer than 5 validation images for one label. Return to the dataset page, add enough validation images, and create a new training task.
+
+
+
+### Deploy to MaixCAM / MaixCAM2
+
+After training is complete and the validation result is acceptable, open the project deployment page and select the training record to deploy. Choose **Manual Deployment**, then click **Download Model** to download the model package.
-For object detection models, after uploading, you need to manually annotate each image by marking the coordinates, size, and category of the objects to be recognized.
-This annotation process can also be done offline on your own computer using software like labelimg, then imported into MaixHub using the dataset import feature.
-Utilize shortcuts during annotation to speed up the process. MaixHub will also add more annotation aids and automatic annotation tools in the future (there is already an automatic annotation tool available for videos that you can try).
+
-### Train the Model
+Unzip the downloaded package first. For a MaixCAM2 image detection model, the package usually contains files such as `model_xxx.mud`, `model_xxx_npu.axmodel`, `model_xxx_vnpu.axmodel`, `main.py`, `app.yaml`, and `report.json`. The `.mud` and `.axmodel` files are the model files that need to be uploaded during manual deployment. `main.py` and `app.yaml` can be used as references for the example program or app configuration.
-Select training parameters, choose the corresponding device platform, select maixcam, and wait in the training queue. You can monitor the training progress in real-time and wait for it to complete.
+
-### Deploy the Model
+Open MaixVision and connect to the device. Open **Device File Manager** on the left, then enter `/root/models` on the device. This directory is used to store model files.
-Once training is complete, you can use the deploy function in the MaixHub app on your device to scan a code and deploy.
-The device will automatically download and run the model, storing it locally for future use.
+
-If you find the recognition results satisfactory, you can share the model to the model library with a single click for others to use.
+Click **Upload File**, select `model_xxx.mud`, `model_xxx_npu.axmodel`, and `model_xxx_vnpu.axmodel` from the unzipped folder, and upload them to `/root/models` on the device.
-## How to Use
+
-Please visit [MaixHub](https://maixhub.com) to register an account, then log in. There are video tutorials on the homepage for learning.
+After uploading, run the example program on the device or load the corresponding `.mud` file from your project code. Test the model in the real scene and check whether the detection box position, confidence, and false detections meet expectations.
-Note that if the tutorial uses the M2dock development board, the process is similar for MaixCAM, although the MaixHub application on the device might differ slightly. The overall process is the same, so please apply the knowledge flexibly.
+
diff --git a/docs/doc/zh/README.md b/docs/doc/zh/README.md
index 6f285e89..914ec993 100644
--- a/docs/doc/zh/README.md
+++ b/docs/doc/zh/README.md
@@ -14,6 +14,41 @@ title: MaixCAM MaixPy 快速开始
border-radius: 0.5em;
border: 1em solid white;
}
+ .device-list {
+ display: flex;
+ flex-direction: column;
+ gap: 0.8em;
+ margin: 1.2em 0;
+ }
+ .device-item {
+ display: grid;
+ grid-template-columns: 13em 1fr auto;
+ gap: 1em;
+ align-items: center;
+ border: 1px solid #e6e8ef;
+ border-radius: 8px;
+ padding: 0.8em;
+ background: #fff;
+ }
+ .device-item img {
+ width: 100%;
+ height: 6.8em;
+ object-fit: contain;
+ background: #f7f8fb;
+ border-radius: 6px;
+ }
+ .device-name {
+ display: block;
+ font-weight: 700;
+ font-size: 1.05em;
+ }
+ .device-desc {
+ margin: 0;
+ }
+ .device-link {
+ white-space: nowrap;
+ font-weight: 600;
+ }
@media screen and (max-width: 900px){
#head_links th, #head_links td {
@@ -21,6 +56,15 @@ title: MaixCAM MaixPy 快速开始
font-size: 0.9em;
padding: 0.1em 0.05em;
}
+ .device-item {
+ grid-template-columns: 1fr;
+ }
+ .device-item img {
+ height: 8em;
+ }
+ .device-link {
+ white-space: normal;
+ }
}
@@ -50,12 +94,27 @@ title: MaixCAM MaixPy 快速开始
> 关于 MaixPy 介绍请看 [MaixPy 官网首页](../../README.md)
> 喜欢 MaixPy 请给 [ MaixPy 项目](https://github.com/sipeed/MaixPy) 点个 Star ⭐️ 以鼓励我们开发更多功能。
-## 快速预览
-
-### Train the Model
+Unzip the downloaded package first. For a MaixCAM2 image detection model, the package usually contains files such as `model_xxx.mud`, `model_xxx_npu.axmodel`, `model_xxx_vnpu.axmodel`, `main.py`, `app.yaml`, and `report.json`. The `.mud` and `.axmodel` files are the model files that need to be uploaded during manual deployment. `main.py` and `app.yaml` can be used as references for the example program or app configuration.
-Select training parameters, choose the corresponding device platform, select maixcam, and wait in the training queue. You can monitor the training progress in real-time and wait for it to complete.
+
-### Deploy the Model
+Open MaixVision and connect to the device. Open **Device File Manager** on the left, then enter `/root/models` on the device. This directory is used to store model files.
-Once training is complete, you can use the deploy function in the MaixHub app on your device to scan a code and deploy.
-The device will automatically download and run the model, storing it locally for future use.
+
-If you find the recognition results satisfactory, you can share the model to the model library with a single click for others to use.
+Click **Upload File**, select `model_xxx.mud`, `model_xxx_npu.axmodel`, and `model_xxx_vnpu.axmodel` from the unzipped folder, and upload them to `/root/models` on the device.
-## How to Use
+
-Please visit [MaixHub](https://maixhub.com) to register an account, then log in. There are video tutorials on the homepage for learning.
+After uploading, run the example program on the device or load the corresponding `.mud` file from your project code. Test the model in the real scene and check whether the detection box position, confidence, and false detections meet expectations.
-Note that if the tutorial uses the M2dock development board, the process is similar for MaixCAM, although the MaixHub application on the device might differ slightly. The overall process is the same, so please apply the knowledge flexibly.
+
diff --git a/docs/doc/zh/README.md b/docs/doc/zh/README.md
index 6f285e89..914ec993 100644
--- a/docs/doc/zh/README.md
+++ b/docs/doc/zh/README.md
@@ -14,6 +14,41 @@ title: MaixCAM MaixPy 快速开始
border-radius: 0.5em;
border: 1em solid white;
}
+ .device-list {
+ display: flex;
+ flex-direction: column;
+ gap: 0.8em;
+ margin: 1.2em 0;
+ }
+ .device-item {
+ display: grid;
+ grid-template-columns: 13em 1fr auto;
+ gap: 1em;
+ align-items: center;
+ border: 1px solid #e6e8ef;
+ border-radius: 8px;
+ padding: 0.8em;
+ background: #fff;
+ }
+ .device-item img {
+ width: 100%;
+ height: 6.8em;
+ object-fit: contain;
+ background: #f7f8fb;
+ border-radius: 6px;
+ }
+ .device-name {
+ display: block;
+ font-weight: 700;
+ font-size: 1.05em;
+ }
+ .device-desc {
+ margin: 0;
+ }
+ .device-link {
+ white-space: nowrap;
+ font-weight: 600;
+ }
@media screen and (max-width: 900px){
#head_links th, #head_links td {
@@ -21,6 +56,15 @@ title: MaixCAM MaixPy 快速开始
font-size: 0.9em;
padding: 0.1em 0.05em;
}
+ .device-item {
+ grid-template-columns: 1fr;
+ }
+ .device-item img {
+ height: 8em;
+ }
+ .device-link {
+ white-space: normal;
+ }
}
@@ -50,12 +94,27 @@ title: MaixCAM MaixPy 快速开始
> 关于 MaixPy 介绍请看 [MaixPy 官网首页](../../README.md)
> 喜欢 MaixPy 请给 [ MaixPy 项目](https://github.com/sipeed/MaixPy) 点个 Star ⭐️ 以鼓励我们开发更多功能。
-## 快速预览
-
+ 
+ 
+