any support for --grid parameter while exporting .onnx model? about yolov7_openvino_cpp-python HOT 24 CLOSED

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

from yolov7_openvino_cpp-python.

hi @akashAD98 We have this notebook to demonstrate the post-processing with --grid parameter
https://github.com/openvinotoolkit/openvino_notebooks/blob/main/notebooks/226-yolov7-optimization/226-yolov7-optimization.ipynb

from yolov7_openvino_cpp-python.

Which means you can feed the model's single output directly into nms module without concatenating 3 of them.

from yolov7_openvino_cpp-python.

@OpenVINO-dev-contest Yes i tried that repo but im facing issues for inference on video/webcams? So I want to use your code for webcams/video inference

from yolov7_openvino_cpp-python.

without grid im getting no detection here

from yolov7_openvino_cpp-python.

also I'm not able to read my custom_model ,which has without --grid parameter

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-13-db161e3ad74f> in <module>
      2 core = Core()
      3 # read converted model
----> 4 model = core.read_model('model/best_veh_withbgnew.xml')
      5 # load model on CPU device
      6 compiled_model = core.compile_model(model, 'CPU')

RuntimeError: Check 'false' failed at C:\Jenkins\workspace\private-ci\ie\build-windows-vs2019\b\repos\openvino\src\frontends\common\src\frontend.cpp:54:
Converting input model
Incorrect weights in bin file!

from yolov7_openvino_cpp-python.

also I'm not able to read my custom_model ,which has without --grid parameter

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-13-db161e3ad74f> in <module>
      2 core = Core()
      3 # read converted model
----> 4 model = core.read_model('model/best_veh_withbgnew.xml')
      5 # load model on CPU device
      6 compiled_model = core.compile_model(model, 'CPU')

RuntimeError: Check 'false' failed at C:\Jenkins\workspace\private-ci\ie\build-windows-vs2019\b\repos\openvino\src\frontends\common\src\frontend.cpp:54:
Converting input model
Incorrect weights in bin file!

Did you get the model from model optimizer. and error happened during model offline converting ?

from yolov7_openvino_cpp-python.

without grid im getting no detection here

Yes this notebook is only for grid model

from yolov7_openvino_cpp-python.

also I'm not able to read my custom_model ,which has without --grid parameter

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-13-db161e3ad74f> in <module>
      2 core = Core()
      3 # read converted model
----> 4 model = core.read_model('model/best_veh_withbgnew.xml')
      5 # load model on CPU device
      6 compiled_model = core.compile_model(model, 'CPU')

RuntimeError: Check 'false' failed at C:\Jenkins\workspace\private-ci\ie\build-windows-vs2019\b\repos\openvino\src\frontends\common\src\frontend.cpp:54:
Converting input model
Incorrect weights in bin file!

Did you get the model from model optimizer. and error happened during model offline converting ?

Im using

from openvino.tools import mo
from openvino.runtime import serialize

model = mo.convert_model('model/best_veh_withbgnew.onnx')
# serialize model for saving IR
serialize(model, 'model/best_veh_withbgnew.xml')

for conversion, .XML file is stored in system ,but not able to read it

from yolov7_openvino_cpp-python.

without grid im getting no detection here

Yes this notebook is only for grid model

1.can you please give some suggestions regarding how can I use --grid in your yolov.py code?
2.in order to convert .XML into int8 format using NNCF ,should I need to pass data? for custom model what kind of data format I need to give? & it requires data in annotation format?(yolo format can I directly use (images & .txt format))

from yolov7_openvino_cpp-python.

without grid im getting no detection here

Yes this notebook is only for grid model

1.can you please give some suggestions regarding how can I use --grid in your yolov.py code? 2.in order to convert .XML into int8 format using NNCF ,should I need to pass data? for custom model what kind of data format I need to give? & it requires data in annotation format?(yolo format can I directly use (images & .txt for

also I'm not able to read my custom_model ,which has without --grid parameter

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
<ipython-input-13-db161e3ad74f> in <module>
      2 core = Core()
      3 # read converted model
----> 4 model = core.read_model('model/best_veh_withbgnew.xml')
      5 # load model on CPU device
      6 compiled_model = core.compile_model(model, 'CPU')

RuntimeError: Check 'false' failed at C:\Jenkins\workspace\private-ci\ie\build-windows-vs2019\b\repos\openvino\src\frontends\common\src\frontend.cpp:54:
Converting input model
Incorrect weights in bin file!

Did you get the model from model optimizer. and error happened during model offline converting ?

Im using

from openvino.tools import mo
from openvino.runtime import serialize

model = mo.convert_model('model/best_veh_withbgnew.onnx')
# serialize model for saving IR
serialize(model, 'model/best_veh_withbgnew.xml')

for conversion, .XML file is stored in system ,but not able to read it

Did you got a .bin file in the same fold with .xml file ?

from yolov7_openvino_cpp-python.

@OpenVINO-dev-contest yes i got .bin & .xml file in folder

from yolov7_openvino_cpp-python.

@OpenVINO-dev-contest issue has been solved, i restarted my system & it solved issue.

i have another question for NNCF Post-training Quantization ,for custom model what format data I need to pass? should I keep val2017 coco data or my own data ?

my goal is to convert into int8 format & without data I think its not possible

quantized_model = nncf.quantize(model, quantization_dataset, preset=nncf.QuantizationPreset.MIXED)

serialize(quantized_model, 'model/yolov7-tiny_int8.xml')

from yolov7_openvino_cpp-python.

@OpenVINO-dev-contest issue has been solved, i restarted my system & it solved issue.

i have another question for NNCF Post-training Quantization ,for custom model what format data I need to pass? should I keep val2017 coco data or my own data ?

my goal is to convert into int8 format & without data I think its not possible

quantized_model = nncf.quantize(model, quantization_dataset, preset=nncf.QuantizationPreset.MIXED)

serialize(quantized_model, 'model/yolov7-tiny_int8.xml')

You should define your dataloader and preprocessing firstly. In notebook example, we use COCO format.

from yolov7_openvino_cpp-python.

from yolov7_openvino_cpp-python.

@bbartling i was able to run both ob window & ubantu linux system

from yolov7_openvino_cpp-python.

from yolov7_openvino_cpp-python.

size in term of wh or model memory size? its 640640 & yolov7-tiny.onnx & yolov7-tiny.bin is 24mb

from yolov7_openvino_cpp-python.

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

from yolov7_openvino_cpp-python.

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

updated, you can add a true after the cpp running command

from yolov7_openvino_cpp-python.

Thank you so much!

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

updated, you can add a true after the cpp running command

from yolov7_openvino_cpp-python.

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

updated, you can add a true after the cpp running command

Just one more question, what's total_num = 25200 means? My custom model only have 10 classes instead of 85. I think I also need to change this number along with 85 to 15 in the code?

from yolov7_openvino_cpp-python.

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

updated, you can add a true after the cpp running command

Just one more question, what's total_num = 25200 means? My custom model only have 10 classes instead of 85. I think I also need to change this number along with 85 to 15 in the code?

25200 is the maximum number of objects model can detect, and yes should switch 85 to 15. pls ensure your model output shape is like [1, 25200, 15] before you change the code.

from yolov7_openvino_cpp-python.

Hi @akashAD98 the code is updated, you can trigger the grid model by add parameter --grid True for inference script

Hello, could also you add the grid option for C++ script?

updated, you can add a true after the cpp running command

Just one more question, what's total_num = 25200 means? My custom model only have 10 classes instead of 85. I think I also need to change this number along with 85 to 15 in the code?

25200 is the maximum number of objects model can detect, and yes should switch 85 to 15. pls ensure your model output shape is like [1, 25200, 15] before you change the code.

I got it. Thank you

from yolov7_openvino_cpp-python.