TDA4VM: Porting ADAS application into TDA4VM

Hello;

Thanks for the question.

For video pipeline application development, please refer to this link. 

https://software-dl.ti.com/jacinto7/esd/processor-sdk-rtos-jacinto7/latest/exports/docs/vision_apps/docs/user_guide/BUILD_INSTRUCTIONS.html#BUILD_SOURCE

You can build the demo examples that come with SDK first. Depend on your application, you can modify the examples accordingly.

For Tensorflow; please refer to this link:

https://github.com/TexasInstruments/edgeai-tidl-tools

You can git-clone the TIDL edgeAI tool. 

We recommend to use ONNX format, then import your ONNX format model to TIDL flow.

If you already have a TI EVM, your can verify the model inference on the EVM in real time, and test the vision apps in real time as well, including camera/image-sensor and display in your pipeline.

Best regards

Wen Li

Hello,

Thanks for the reply.

In the edge-AI-tidl sdk, I can see that scripts are available to compile some pre-configured models. Can you please guide to compile a custom onnx model? 

Hello,

The engineer assigned is currently out of office. They will return beginning of next week and will be able to update. We appreciate your patience.

Warm regards,

Christina

For the model compiling guideline:

1. Please create a model instantiation in the "model_configs.py"; this is the file in the downloaded edgeai-TIDL-TOOLS folder, under the "../examples/osrt_python/" sub-folder.
2. I am using the yolo model as an examples and paste below. You should config your file with the parameters that fit your model&Application
3. There is another "common_utils.py" file, please adjust the parameters according to your model as well.

Then you can compile or inference your model.

Here are the commands to compile and inference a model

1. Compile and inference a model
python3 onnxrt_ep.py -c -m yolox_s_pose_ti_lite

2. Inference a model only
python3 onnxrt_ep.py -m yolox_s_pose_ti_lite

=======================================================================================

"od-8200_onnxrt_coco_edgeai-mmdet_yolox_nano_lite_416x416_20220214_model_onnx": create_model_config(
task_type="detection",
source=dict(
model_url="">software-dl.ti.com/.../yolox_nano_lite_416x416_20220214_model.onnx",
meta_arch_url="">software-dl.ti.com/.../yolox_nano_lite_416x416_20220214_model.prototxt",
infer_shape=True,
),
preprocess=dict(
resize=416,
crop=416,
data_layout="NCHW",
pad_color=[114, 114, 114],
resize_with_pad=[True, "corner"],
reverse_channels=True,
),
session=dict(
session_name="onnxrt",
model_path=os.path.join(
models_base_path, "yolox_nano_lite_416x416_20220214_model.onnx"
),
meta_layers_names_list=os.path.join(
models_base_path, "yolox_nano_lite_416x416_20220214_model.prototxt"
),
meta_arch_type=6,
input_mean=[0, 0, 0],
input_scale=[1, 1, 1],
input_optimization=True,
),
postprocess=dict(
formatter="DetectionBoxSL2BoxLS",
resize_with_pad=True,
keypoint=False,
object6dpose=False,
normalized_detections=False,
shuffle_indices=None,
squeeze_axis=None,
reshape_list=[(-1, 5), (-1, 1)],
ignore_index=None,
),
extra_info=dict(
od_type="SSD",
framework="MMDetection",
num_images=numImages,
num_classes=91,
label_offset_type="80to90",
label_offset=1,