Quick Experience

Model Benchmark

After completing the first two steps, the model benchmarking tool axcl_run_model can be used. This tool has many parameters, which can be viewed with axcl_run_model --help. If you are interested in its implementation mechanism, you can check the source code in the corresponding sample directory. This tool, along with other cv & llm sample, is provided in source code form to help users understand how to use the API.

For example, to test the running speed of a model, use a command like axcl_run_model -m your_model.axmodel -r 10, where -m specifies the model to run, and -r specifies the number of repetitions, allowing for a simple test of the model's speed.

Obtain the model:

wget https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/linux/ax8850_card/yolo11s.axmodel

Test:

axcl_run_model -m yolo11s.axmodel -r 10

The execution result is as follows:

m5stack@raspberrypi5:~ $ axcl_run_model -m yolo11s.axmodel -r 10
   Run AxModel:
         model: yolo11s.axmodel
          type: 3 Core
          vnpu: Disable
        warmup: 1
        repeat: 10
         batch: { auto: 1 }
    axclrt ver: 1.0.0
   pulsar2 ver: 3.2 99cf147d
      tool ver: 0.0.1
      cmm size: 10488066 Bytes
  ------------------------------------------------------
  min =   3.391 ms   max =   3.414 ms   avg =   3.402 ms
  ------------------------------------------------------

From the above example, in addition to indicating the model running time, it also shows the toolchain version, model type, and other related information.

CV Example

Note

If the unzip tool is not installed, please install it first with the following command:

sudo apt install zip

Get the Demo:

wget https://m5stack.oss-cn-shenzhen.aliyuncs.com/resource/linux/ax8850_card/cv_demo.zip

Unzip the Demo:

unzip cv_demo.zip

Enter the directory:

cd cv_demo

Classification Model

Using the imagenet dataset’s imagenet_cat.jpg as the classification target, after sample execution, the output will be as follows (note that the model and input image should be adjusted according to the actual situation):

Execution command:

./axcl_sample_classification -m mobilenetv2.axmodel -i cat.jpg

Result:

m5stack@raspberrypi5:~/cv_demo $ ./axcl_sample_classification -m mobilenetv2.axmodel -i cat.jpg
axcl initializing...
axcl inited.
Select axcl device{index: 0} as {1}.
axclrt Engine inited.
--------------------------------------
model file : mobilenetv2.axmodel
image file : cat.jpg
img height : 224
img width  : 224
--------------------------------------
282:  9.8%,  tiger cat
285:  9.8%,  Egyptian cat
283:  9.5%,  Persian cat
281:  9.4%,  tabby, tabby cat
463:  7.5%,  bucket, pail
--------------------------------------

Detection Model

Using the PASCAL VOC dataset’s voc_dog.jpg as the detection target, after sample execution, the output will be as follows (note that the model and input image should be adjusted according to the actual situation):

Execution command:

./axcl_sample_yolov5s -m yolov5s.axmodel -i dog.jpg

Result:

m5stack@raspberrypi5:~/cv_demo $ ./axcl_sample_yolov5s -m yolov5s.axmodel -i dog.jpg
axcl initializing...
axcl inited.
Select axcl device{index: 0} as {1}.
axclrt Engine inited.
--------------------------------------
model file : yolov5s.axmodel
image file : dog.jpg
img height : 640
img width  : 640
--------------------------------------
post process cost time:0.61 ms
--------------------------------------
Repeat 1 times, avg time 8.10 ms, max_time 8.10 ms, min_time 8.10 ms
--------------------------------------
16:  91%, [ 138,  218,  310,  541], dog
 2:  69%, [ 470,   76,  690,  173], car
 1:  56%, [ 158,  120,  569,  420], bicycle

It can be seen that 3 objects were detected and their class ID, confidence, and coordinates were provided. In the sample execution directory, a detection result named yolov5s_out.jpg will be saved, which can be opened with an image viewer to preview the output.

Next Overview

Linux PC

CM4Stack

CoreMP135

AI Accelerator Card

LLM-8850 Card

Quick Start

Vision Models

Large Language Models

Multimodal Models

Audio Models

Generative Models

Advanced Usage

LLM

Real-Time AI Voice Assistant

OpenAI Voice Assistant

XiaoZhi Voice Assistant

XiaoLing Voice Assistant

AtomS3R-M12 Volcengine Kit

Offline Voice Recognition

Unit ASR

Home Assistant

Industrial Control

StamPLC

IoT Measuring Instruments

Air Quality

Module13.2 PPS

VAMeter

T-Lite

Ezdata

Ethernet Camera

PoECAM

Wi-Fi Camera

TimerCAM

Unit CamS3/-5MP

AI Camera

UnitV2

M5StickV/UnitV

LoRa & LoRaWAN

TTN (The Things Network)

Meshtastic

Motor Control

Unit Roller485/CAN

Develop Tools

Hobby Kit

Restore Factory Firmware