when does it arrive?
Hi @MJK , there was an issues of the conversion earlier, now the AI model conversion server is back, you may try to conversion your model again. Thank you.
can you teach me how to convert
im running modified yolov7 tiny pytorch
can you please help me
im not sure if this is correct? and inside the zip is just “best.pt” from after training
and its 6MB
i just got it!
Thank you !! hope this works
Hi @MJK, for yolov7 there is an additional step before using the AI model conversion.
First you need to reparameterization the trained pytouch file (.py) by putting the .py in your Arduino environment in your pc “C:\Users\
{username}\AppData\Local\Arduino15\packages\realtek\hardware\AmebaPro2\4.0.9-build20250325\libraries\NeuralNetwork\src\Yolov7_reparam_scripts”.
Second, Follow this guide to find the command for reparameterization.
Third, after reparameterization then use to AI model conversion to convert into network binary file (.nb) to use it in Ameba pro2. Thank you.
okay ill follow it the reparametrization should be in the repo github right?
hi @MJK , you will found it in you pc Arduino environment “C:\Users
{username}\AppData\Local\Arduino15\packages\realtek\hardware\AmebaPro2\4.0.9-build20250325\libraries\NeuralNetwork\src\Yolov7_reparam_scripts”.
i have found it .
do i need to copy paste my best.pt to do this script?
still errors
hi @MJK , i see that model is missing, you can try download from this repo then place the reparam_yolov7-tiny.py in the downloaded repo root directory, then run again the this “python reparam_yolov7-tiny.py –weights weights/best.pt –custom_yaml custom/yolov7-tiny-deploy.yaml –output best_reparam.pt”. Thank you.
okay bro thank you ill try it ,
I tried to run .nb file without the reparam process it goes like crazy in detection is that possible ?
do i neeed to do this process even though the best.pt is lower than 35MB?
Hi @MJK,
Any yolov7-tiny model trained via pytorch need to go through this reparameterisation process.
Thank you.
I have already download the repo what specific folder should i move the reparam py script? and where would i put weights i trained?
Hi @MJK,
Maybe you can give it a try at your current snapshot location. Did you face any issue while running the command here?
Thank you.
yes
here it says some syntaxes but sill not working
i tried the command provided from the guides in Object Detection
Hi @MJK,
Try with --weights, --custom_yaml with the correct path to the weights and custom_yaml.
Two dashes with no space.
Also, you may refer to these posts, these users have managed to convert successfully before. (Converting .pt to .nb - #3 by Steph_B) (AMB82-mini YOLOv7 Pothole Detection).
Thank you.
python reparam_yolov7-tiny.py --weights ‘weights/best.pt’, --custom_yaml yolov7-tiny-deploy.yaml, –-output best_reparam.pt
i put the best.pt in weights folders inside the yolov7-main
the custom_yaml just also outside with the subfolders
okay bro ill check it
thank youu
side note: i successfully convereted into .nb but i havent reparam , thats why the detection went crazy hehe
sorry for the bother hehe ithink last one?