nebula_moose_12014

Thanks i will check and let you know

I'm using the following environment:

Board: RDK X5
Docker: openexplorer/ai_toolchain_ubuntu_20_x5_cpu:v1.2.8
hb_mapper: 1.24.3
HBDK: 3.49.15
Ultralytics: 8.4.114
Model: YOLOv10n
Export: ONNX (opset 11, simplify=True)
Input: 640×640
Output: (1, 300, 6)

convert.yaml
model_parameters:
onnx_model: "/workspace/Downloads/PhysicalAI/yolov10n.onnx"
march: "bayes-e"
working_dir: "./output"
output_model_file_prefix: "yolov10n_int8"

input_parameters:
input_name: "images"
input_type_train: "rgb"
input_type_rt: "nv12"
input_layout_train: "NCHW"
norm_type: "data_scale"
scale_value: "255"
input_shape: "1x3x640x640"

calibration_parameters:
calibration_type: "default"
preprocess_on: true
cal_data_dir: "/workspace/Downloads/calibration"

compiler_parameters:
optimize_level: "O3"
jobs: 8
compile_mode: "latency"

I am using 100 COCO validation images for calibration.

The model is parsed correctly, calibration starts successfully, but during calibration I get the following ONNX Runtime error:

The input tensor cannot be reshaped to the requested shape.

Input shape:
{8,256,20,20}

Requested shape:
{1,2,128,400}

I also inspected the exported ONNX model. The last operators are:

Concat
Transpose
Split
ReduceMax
TopK
Unsqueeze
Expand
GatherElements
Flatten
TopK
Div
Unsqueeze
Mod
Unsqueeze
Unsqueeze
Cast
GatherElements
Expand
GatherElements
Concat

Has anyone successfully converted an Ultralytics YOLOv10n ONNX model to a .bin for the RDK X5?

RDK X5: hb_mapper fails to compile Ultralytics YOLOv10n ONNX (Reshape error during calibration)

I am trying to connect OLED with kit , i connected everything properly. but it not detecting... did anybody try any sensor/input are working??