You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: README.md
+12-7Lines changed: 12 additions & 7 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,14 @@
1
1
# CMSIS-MLEK Software Pack
2
2
3
-
This repository contains **Machine Learning Evaluation Kit (MLEK)** pack which provides [CMSIS Reference Applications](https://open-cmsis-pack.github.io/cmsis-toolbox/ReferenceApplications/) and templates for Edge AI development. These applications implement data preprocessing, memory management, and neural network inference pipelines that are optimized for Cortex-M and Ethos-U platforms.
3
+
This repository contains **Machine Learning Evaluation Kit (MLEK)** pack which provides [CMSIS Reference Applications](https://open-cmsis-pack.github.io/cmsis-toolbox/ReferenceApplications/) and templates for Edge AI development. These applications implement data preprocessing, memory management, and neural network inference pipelines that are optimized for Cortex-M and Ethos-U platforms. The examples are prepared to run with [Keil Studio for VS Code](https://www.keil.arm.com/).
4
+
5
+
The CMSIS-MLEK software pack is derived from the [Arm® ML embedded evaluation kit](https://git.gitlab.arm.com/artificial-intelligence/ethos-u/ml-embedded-evaluation-kit) and makes the examples easier to access. It also contains interfaces to physical hardware and simplifies porting to target hardware. It contains the following ML applications and uses currently Neural Network Models currently in [TensorFlow Lite](https://www.keil.arm.com/packs/tensorflow-lite-micro-tensorflow) format.
6
+
7
+
ML application | Description | Neural Network Model
[Keyword spotting (KWS)](./template/audio) | Recognize the presence of a key word in verbal speech | [MicroNet](https://github.com/ARM-software/ML-zoo/tree/9f506fe52b39df545f0e6c5ff9223f671bc5ae00/models/keyword_spotting/micronet_medium/tflite_int8)
10
+
[Object detection](./template/video) | Detects and draws face bounding box in a given image | [Yolo Fastest](https://github.com/emza-vs/ModelZoo/blob/master/object_detection/yolo-fastest_192_face_v4.tflite)
11
+
[Generic inference runner](./template/generic) | Code block allowing you to develop your own use case | Your custom model
4
12
5
13
Refer to [Overview](./overview/README.md) for more details.
6
14
@@ -9,14 +17,15 @@ Refer to [Overview](./overview/README.md) for more details.
[overview](./overview) | Top-level overview of MLEK reference applications.
12
-
[documentation](./documentation/)| [User documentation](https://arm-examples.github.io/cmsis-mlek/main/index.html) of the MLEK reference applications.
20
+
[docs](./docs/) | Source of the [user documentation](https://arm-examples.github.io/cmsis-mlek) of the MLEK reference applications.
13
21
[template](./template) | [MLEK reference applications](https://arm-examples.github.io/cmsis-mlek/main/template.html), a test framework for DSP and ML algorithms.
14
22
[.github/workflows](./.github/workflows) | GitHub Actions for validation and publishing.
15
23
[.ci](./.ci) | Files that relate to CI tests.
16
24
17
25
## Related
18
26
19
27
-[MLEK Pack](https://www.keil.arm.com/packs/MLEK) available on [keil.arm.com/packs](https://www.keil.arm.com/packs).
28
+
-[tensorflow-lite-micro Pack](https://www.keil.arm.com/packs/tensorflow-lite-micro-tensorflow) available on [keil.arm.com/packs](https://www.keil.arm.com/packs).
20
29
-[MLEK Examples](https://github.com/Arm-Examples/mlek-examples) a repository with MLEK applications that are configured for evaluation boards and use cases.
21
30
-[ML Developers Guide for Cortex-M Processors and Ethos-U NPU](https://developer.arm.com/documentation/109267).
22
31
-[Arm Virtual Hardware - FVP](https://github.com/arm-software/avh) repository with documentation.
@@ -27,11 +36,7 @@ The MLEK references applications are licensed under [ is generated using [MKDocs](https://www.mkdocs.org/) with the following additional plugins:
31
-
32
-
-[mermaid2](https://mkdocs-mermaid2.readthedocs.io/en/latest/) for sequence diagrams.
33
-
34
-
Use `mkdocs serve` to generate the documentation on a local computer.
39
+
The [documentation](https://arm-examples.github.io/cmsis-mlek/index.html) is generated using [MKDocs](https://www.mkdocs.org/). Use `mkdocs serve` to generate the documentation on a local computer.
Copy file name to clipboardExpand all lines: docs/overview.md
+18-15Lines changed: 18 additions & 15 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -10,14 +10,15 @@ Key Features:
10
10
-**Performance Validation**: Test and optimize ML performance on target hardware or simulation.
11
11
-**Hardware Evaluation**: Compare performance across different Corstone platforms and configurations.
12
12
13
-
MLEK covers three main application domains:
13
+
The CMSIS-MLEK software pack is derived from the [Arm® ML embedded evaluation kit](https://git.gitlab.arm.com/artificial-intelligence/ethos-u/ml-embedded-evaluation-kit) and makes the examples easier to access. It also contains interfaces to physical hardware and simplifies porting to target hardware. It contains the following ML applications and uses currently Neural Network Models currently in [TensorFlow Lite](https://www.keil.arm.com/packs/tensorflow-lite-micro-tensorflow) format.
14
14
15
-
-[Audio](https://arm-examples.github.io/cmsis-mlek/templates_audio.html) processing with audio classification and keyword spotting.
16
-
-[Video](https://arm-examples.github.io/cmsis-mlek/templates_video.html) processing with image classification and object detection.
17
-
-[Generic](https://arm-examples.github.io/cmsis-mlek) inference for flexible ML model deployment.
15
+
| ML application | Description | Neural Network Model |
|[Keyword spotting (KWS)](https://github.com/ARM-software/cmsis-mlek/template/audio)| Recognize the presence of a key word in verbal speech |[MicroNet](https://github.com/ARM-software/ML-zoo/tree/9f506fe52b39df545f0e6c5ff9223f671bc5ae00/models/keyword_spotting/micronet_medium/tflite_int8)|
18
+
|[Object detection](https://github.com/ARM-software/cmsis-mlek/template/video)| Detects and draws face bounding box in a given image |[Yolo Fastest](https://github.com/emza-vs/ModelZoo/blob/master/object_detection/yolo-fastest_192_face_v4.tflite)|
19
+
|[Generic inference runner](https://github.com/ARM-software/cmsis-mlek/template/generic)| Code block allowing you to develop your own use case | Your custom model |
18
20
19
-
20
-
Each reference applications is a *csolution project* which supports deployment to physical hardware or Arm Virtual Hardware (AVH-FVP) for simulation.
21
+
Each [ML reference application](https://open-cmsis-pack.github.io/cmsis-toolbox/ReferenceApplications/) is a *csolution project* which supports deployment to physical hardware or Arm Virtual Hardware (AVH-FVP) for simulation.
@@ -27,24 +28,26 @@ A board layer (`*.clayer.yml`) implements the drivers for the physical interface
27
28
|:----------------------------|:----------------|
28
29
|**Audio Processing**||
29
30
| CMSIS_VSTREAM_AUDIO_IN |[CMSIS-Driver vStream](https://arm-software.github.io/CMSIS_6/latest/Driver/group__mci__interface__gr.html) configured for Audio input. |
31
+
| STDOUT | Standard I/O for printf output. |
30
32
|**Video Processing**||
31
33
| CMSIS_VSTREAM_VIDEO_IN |[CMSIS-Driver vStream](https://arm-software.github.io/CMSIS_6/latest/Driver/group__mci__interface__gr.html) configured for Video input. |
32
34
| CMSIS_VSTREAM_VIDEO_OUT |[CMSIS-Driver vStream](https://arm-software.github.io/CMSIS_6/latest/Driver/group__mci__interface__gr.html) configured for Video output. |
35
+
| STDOUT | Standard I/O for printf output. |
33
36
|**Generic Inference Runner**||
34
-
|CMSIS_VSTREAM_AUDIO_IN|[CMSIS-Driver vStream](https://arm-software.github.io/CMSIS_6/latest/Driver/group__mci__interface__gr.html configured for Audio input. |
37
+
|STDOUT| Standard I/O for printf output. |
35
38
36
39
## Platform Support
37
40
38
-
The templates support via _target names_ multiple [Arm Cortex-M IP Subsystems](https://www.arm.com/products/silicon-ip-subsystems#Products).
41
+
The templates support via _target names_ multiple [Arm Cortex-M IP Subsystems](https://www.arm.com/products/silicon-ip-subsystems#Products). These _target names_ support execution on [AVH FVP simulation models](https://arm-software.github.io/AVH/main/simulation/html/index.html) which is useful during software development or with [Contiguous Integration (CI)](https://github.com/Arm-Examples/cmsis-mlek/actions) testing using GitHub actions.
39
42
40
-
| IP Subsystem | Description |
41
-
|:-------------|:-------------|
42
-
| Corstone-300 | Cortex-M55 optional with Ethos-U55 or Ethos-U65 |
43
-
| Corstone-310 | Cortex-M85 optional with Ethos-U55 |
44
-
| Corstone-315 | Cortex-M85 optional with Ethos-U65 |
45
-
| Corstone-320 | Cortex-M85 optional with Ethos-U85 |
43
+
|Target Name |IP Subsystem | Description |
44
+
|:-------------|:-------------|:-------------|
45
+
|AVH-SSE-300 |[Corstone-300](https://developer.arm.com/Processors/Corstone-300)| Cortex-M55 optional with Ethos-U55 or Ethos-U65 |
46
+
|AVH-SSE-310 |[Corstone-310](https://developer.arm.com/Processors/Corstone-310)| Cortex-M85 optional with Ethos-U55 |
47
+
|AVH-SSE-315 |[Corstone-315](https://developer.arm.com/Processors/Corstone-315)| Cortex-M85 optional with Ethos-U65 |
48
+
|AVH-SSE-320 |[Corstone-320](https://developer.arm.com/Processors/Corstone-320)| Cortex-M85 optional with Ethos-U85 |
46
49
47
-
Adding a postfix to the _target name_ in the `*.csolution.yml` project file configures the neural network inference pipeline for Ethos-U. Without this prefix only the Cortex-M system is used as shown in the diagram below.
50
+
Adding a postfix to the _target name_ in the `*.csolution.yml` project file configures the neural network inference pipeline for Ethos-U. Without this prefix only the Cortex-M system is used as shown in the diagram below. Such a postfix can also be used for _target names_ that deploy to physical hardware.
0 commit comments