This document describes the process of setting up and running the Arm® Ethos™-U NPU Visual Wake Word example. Visual Wake Words is a common vision use-case to detect if a the provided image contains a person.
Use case code could be found in source/use_case/vww directory.
See Prerequisites
In addition to the already specified build option in the main reference manual, Visual Wake Word use case specifies:
vww_MODEL_TFLITE_PATH - Path to the NN model file in the TFLite format. The model is then processed and included in the application axf file. The default value points to one of the delivered set of models. Note that the parameters vww_LABELS_TXT_FILE, TARGET_PLATFORM, and ETHOS_U_NPU_ENABLED must be aligned with the chosen model. In other words:
ETHOS_U_NPU_ENABLED is set to On or 1, then the NN model is assumed to be optimized. The model naturally falls back to the Arm® Cortex®-M CPU if an unoptimized model is supplied.ETHOS_U_NPU_ENABLED is set to Off or 0, the NN model is assumed to be unoptimized. Supplying an optimized model in this case results in a runtime error.vww_FILE_PATH: Path to directory or file to be used as custom image file(s) to use in the evaluation application. The default value points to the resources/vww/samples folder containing the delivered set of images. See more in the Running custom input data section.
vww_IMAGE_SIZE: The NN model requires input images to be of a specific size. This parameter defines the size of the image side in pixels. Images are considered squared. Default value is 128, which is what the supplied visual wake word model expects.
vww_LABELS_TXT_FILE: Path to the labels' text file to be baked into the application. The file is used to map classified classes index to the text label. Change this parameter to point to the custom labels file to map custom NN model output correctly.
The default value points to the delivered labels.txt file inside the delivery package.
vww_ACTIVATION_BUF_SZ: The intermediate/activation buffer size reserved for the NN model. By default, it is set to 2MiB and should be enough for most models.
Note: This section describes the process for configuring the build for
MPS3: SSE-300for different target platform see Building section.
Create a build directory and navigate inside:
mkdir build_visual_wake_word && cd build_visual_wake_word
On Linux, execute the following command to build only Visual Wake Word application to run on the Ethos-U55 Fast Model when providing only the mandatory arguments for CMake configuration:
cmake ../ -DUSE_CASE_BUILD=vww
To configure a build that can be debugged using Arm-DS, we can just specify the build type as Debug and use the Arm Compiler toolchain file:
cmake .. \ -DCMAKE_TOOLCHAIN_FILE=scripts/cmake/toolchains/bare-metal-armclang.cmake \ -DCMAKE_BUILD_TYPE=Debug \ -DUSE_CASE_BUILD=vww
Also see:
Note: If re-building with changed parameters values, it is highly advised to clean the build directory and re-run the CMake command.
If the CMake command succeeded, build the application as follows:
make -j4
Add VERBOSE=1 to see compilation and link details.
Results of the build will be placed under build/bin folder:
bin ├── ethos-u-vww.axf ├── ethos-u-vww.htm ├── ethos-u-vww.map ├── images-vww.txt └── sectors └── vww ├── ddr.bin └── itcm.bin
Where:
ethos-u-vww.axf: The built application binary for the Visual Wake Word use case.
ethos-u-vww.map: Information from building the application (e.g. libraries used, what was optimized, location of objects)
ethos-u-vww.htm: Human readable file containing the call graph of application functions.
sectors/: Folder containing the built application, split into files for loading into different FPGA memory regions.
Images-vww.txt: Tells the FPGA which memory regions to use for loading the binaries in sectors/** folder.
The application performs inference on image data found in the folder set by the CMake parameter vww_FILE_PATH.
To run the application with your own images first create a folder to hold them and then copy the custom images into this folder:
mkdir /tmp/custom_images cp custom_image1.bmp /tmp/custom_images/
Note: Clean the build directory before re-running the cmake command.
Next set vww_FILE_PATH to the location of this folder when building:
cmake .. \ -Dvww_FILE_PATH=/tmp/custom_images/ \ -DUSE_CASE_BUILD=vww
The images found in the vww_FILE_PATH folder will be picked up and automatically converted to C++ files during the CMake configuration stage and then compiled into the application during the build phase for performing inference with.
The log from the configuration stage should tell you what image directory path has been used:
-- User option vww_FILE_PATH is set to /tmp/custom_images -- User option vww_IMAGE_SIZE is set to 128 ... -- Generating image files from /tmp/custom_images ++ Converting custom_image1.bmp to custom_image1.cc ... -- Defined build user options: ... -- vww_FILE_PATH=/tmp/custom_images -- vww_IMAGE_SIZE=128
After compiling, your custom images will have now replaced the default ones in the application.
Note: The CMake parameter vww_IMAGE_SIZE should match the model input size. When building the application, if the size of any image does not match IMAGE_SIZE then it will be rescaled and padded so that it does.
The application performs inference using the model pointed to by the CMake parameter vww_MODEL_TFLITE_PATH.
Note: If you want to run the model using Ethos-U, ensure your custom model has been run through the Vela compiler successfully before continuing.
To run the application with a custom model you will need to provide a labels_<model_name>.txt file of labels associated with the model. Each line of the file should correspond to one of the outputs in your model. See the provided visual_wake_word_labels.txt file for an example.
Then, you must set vww_MODEL_TFLITE_PATH to the location of the Vela processed model file and vww_LABELS_TXT_FILE to the location of the associated labels file.
An example:
cmake \ -Dvww_MODEL_TFLITE_PATH=<path/to/custom_model_after_vela.tflite> \ -Dvww_LABELS_TXT_FILE=<path/to/labels_custom_model.txt> \ -DUSE_CASE_BUILD=vww ..
Note: Clean the build directory before re-running the cmake command.
The TFLite model pointed to by vww_MODEL_TFLITE_PATH and labels text file pointed to by vww_LABELS_TXT_FILE will be converted to C++ files during the CMake configuration stage and then compiled into the application for performing inference with.
The log from the configuration stage should tell you what model path and labels file have been used:
-- User option vww_MODEL_TFLITE_PATH is set to <path/to/custom_model_after_vela.tflite> ... -- User option vww_LABELS_TXT_FILE is set to <path/to/labels_custom_model.txt> ... -- Using <path/to/custom_model_after_vela.tflite> ++ Converting custom_model_after_vela.tflite to custom_model_after_vela.tflite.cc -- Generating labels file from <path/to/labels_custom_model.txt> -- writing to <path/to/build/generated/src/Labels.cc> ...
After compiling, your custom model will have now replaced the default one in the application.
The FVP is available publicly from Arm Ecosystem FVP downloads.
For the Ethos-U evaluation, please download the MPS3 based version of the Arm® Corstone™-300 model that contains Cortex-M55 and offers a choice of the Ethos-U55 and Ethos-U65 processors.
Unpack the archive
Run the install script in the extracted package
$./FVP_Corstone_SSE-300.sh
Pre-built application binary ethos-u-vww.axf can be found in the bin/mps3-sse-300 folder of the delivery package. Assuming the install location of the FVP was set to ~/FVP_install_location, the simulation can be started by:
$ ~/FVP_install_location/models/Linux64_GCC-6.4/FVP_Corstone_SSE-300_Ethos-U55 ./bin/mps3-sse-300/ethos-u-vww.axf
A log output should appear on the terminal:
telnetterminal0: Listening for serial connection on port 5000 telnetterminal1: Listening for serial connection on port 5001 telnetterminal2: Listening for serial connection on port 5002 telnetterminal5: Listening for serial connection on port 5003
This will also launch a telnet window with the sample application's standard output and error log entries containing information about the pre-built application version, TensorFlow Lite Micro library version used, data type as well as the input and output tensor sizes of the model compiled into the executable binary.
After the application has started if vww_FILE_PATH pointed to a single file (or a folder containing a single image) the inference starts immediately. In case of multiple inputs choice, it outputs a menu and waits for the user input from telnet terminal:
User input required Enter option number from: 1. Classify next ifm 2. Classify ifm at chosen index 3. Run classification on all ifm 4. Show NN model info 5. List ifm Choice:
“Classify next image” menu option will run single inference on the next in line image from the collection of the compiled images.
“Classify image at chosen index” menu option will run single inference on the chosen image.
Note: Please make sure to select image index in the range of supplied images during application build. By default, pre-built application has 2 images, index should 0 or 1.
“Run classification on all images” menu option triggers sequential inference executions on all built-in images.
“Show NN model info” menu option prints information about model data type, input and output tensor sizes:
INFO - Added ethos-u support to op resolver INFO - Creating allocator using tensor arena in SRAM INFO - Allocating tensors INFO - Model INPUT tensors: INFO - tensor type is INT8 INFO - tensor occupies 16384 bytes with dimensions INFO - 0: 1 INFO - 1: 128 INFO - 2: 128 INFO - 3: 1 INFO - Quant dimension: 0 INFO - Scale[0] = 0.008138 INFO - ZeroPoint[0] = -70 INFO - Model OUTPUT tensors: INFO - tensor type is INT8 INFO - tensor occupies 2 bytes with dimensions INFO - 0: 1 INFO - 1: 2 INFO - Quant dimension: 0 INFO - Scale[0] = 0.022299 INFO - ZeroPoint[0] = -17 INFO - Activation buffer (a.k.a tensor arena) size used: 133716 INFO - Number of operators: 19 INFO - Operator 0: ethos-u INFO - Operator 1: PAD INFO - Operator 2: ethos-u INFO - Operator 3: PAD INFO - Operator 4: ethos-u INFO - Operator 5: PAD INFO - Operator 6: ethos-u INFO - Operator 7: PAD INFO - Operator 8: ethos-u INFO - Operator 9: PAD INFO - Operator 10: ethos-u INFO - Operator 11: PAD INFO - Operator 12: ethos-u INFO - Operator 13: PAD INFO - Operator 14: ethos-u INFO - Operator 15: PAD INFO - Operator 16: ethos-u INFO - Operator 17: AVERAGE_POOL_2D INFO - Operator 18: ethos-u
“List Images” menu option prints a list of pair image indexes - the original filenames embedded in the application:
INFO - List of Files: INFO - 0 => man_in_red_jacket.png INFO - 1 => st_paul_s_cathedral.png
Please select the first menu option to execute Visual Wake Word.
The following example illustrates application output for classification:
INFO - Running inference on image 0 => man_in_red_jacket.png INFO - Final results: INFO - Total number of inferences: 1 INFO - 0) 1 (3.211100) -> person INFO - Profile for Inference: INFO - NPU AXI0_RD_DATA_BEAT_RECEIVED beats: 228679 INFO - NPU AXI0_WR_DATA_BEAT_WRITTEN beats: 153031 INFO - NPU AXI1_RD_DATA_BEAT_RECEIVED beats: 40625 INFO - NPU ACTIVE cycles: 706754 INFO - NPU IDLE cycles: 10954 INFO - NPU TOTAL cycles: 717708
It could take several minutes to complete one inference run (average time is 2-3 minutes).
The log shows the inference results for “image 1” (1 - index) that corresponds to man_in_red_jacket.png” in the sample image resource folder.
The profiling section of the log shows that for this inference:
Ethos-U's PMU report:
717,708 total cycle: The number of NPU cycles
706,754 active cycles: number of NPU cycles that were used for computation
10,954 idle cycles: number of cycles for which the NPU was idle
228,679 AXI0 read beats: The number of AXI beats with read transactions from AXI0 bus. AXI0 is the bus where Ethos-U NPU reads and writes to the computation buffers (activation buf/tensor arenas).
153,031 AXI0 write beats: The number of AXI beats with write transactions to AXI0 bus.
40,625 AXI1 read beats: The number of AXI beats with read transactions from AXI1 bus. AXI1 is the bus where Ethos-U NPU reads the model (read only)
For FPGA platforms, CPU cycle count can also be enabled. For FVP, however, CPU cycle counters should not be used as the CPU model is not cycle-approximate or cycle-accurate.
The application prints the detection with label index, confidence score and labels from associated pd_labels.txt file.