model already load,but the error is model is not loaded

[C4712]

[kadirnar]

Hi @C4712

Are you working locally? Are you using a custom model? Can you check the yolov7_model_path variable?

[IdrissSinapan974]

Hi !
I have same error, I work locally, I use a custom model and my yolov7_model_path variable is correct (path to the weight I obtained after training).
It worked the last time and now I don't know why It is not the case.
Thank you :)

[kadirnar]

Can you share the error message? Are you using the colab file?

[IdrissSinapan974]

I clone the repo in my computer, because I have to much data to import on colab...

[kadirnar]

Are you using the ipynb file? https://github.com/kadirnar/Yolov7-SAHI/blob/main/demo/inference_for_yolov7.ipynb

[IdrissSinapan974]

Yes, I use the .ipynb file, I copy and past what I need and adapt the lines for my case. (yolov.pt is the weight I obtained after training )

[beratersari]

Same Issue

[IdrissSinapan974]

Resolved !
In model.py file, in "class Yolov7DetectionModel(DetectionModel):" they used torch.hub.load(...) and for this you need internet connection.... For my case, I tested the code, but after that I use it in a computer without internet connection. And this error appear.
Hope it will help.

[beratersari]

I have an internet connection already. Do you use custom model? The original yolov7.pt works for me. But my custom model doesnt work

[IdrissSinapan974]

I did yolov7 training on my custom dataset, so I obtain best.pt. Then I used the obtain weight (best.pt renamed yolov7.pt) in SAHI. Do you have the exact same error (Model is not loaded ) ?

[kadirnar]

@IdrissSinapan974

Does the error go away when I type the filename yolov7.pt?

[IdrissSinapan974]

@kadirnar no, the filename have no effect on this issue (for my case). I changed the filename to something different to "yolov7.pt" and it works too.

[pythoncodeAshisH]

In model.py file, in "class Yolov7DetectionModel(DetectionModel):" they used torch.hub.load(...)

but In my case i have good internet connection still i got error

[IdrissSinapan974]

@pythoncodeAshisH, can you share the code section which gives you this error please ?

[pythoncodeAshisH]

@IdrissSinapan974 Sure....!

how to solve this error

[kadirnar]

@pythoncodeAshisH
Can you share your model? I will update the yolov7-pip package this week. I'll add it to the SAHI algorithm.

Can you try to load your custom model using Yolov7_hubconf?
https://github.com/WongKinYiu/yolov7/blob/main/hubconf.py

This bug is related to the yolov7 repo. To fix this, we can open a Pull Request.

[beratersari]

@pythoncodeAshisH Hey, I am also getting error from the predictor which is on the beginning of the notebook. Use the prediction function which is at the end of the notebook. I shared a screenshot below. Use that

[IdrissSinapan974]

@pythoncodeAshisH Try to do like @beratersari suggested it should work. There is different way to use SAHI (for only one picture, or a folder containing a lot of pictures).
If you absolutely need the structure like you are trying to use I will try to post a screenshot of this case of use.

[pythoncodeAshisH]

I am trying to use the instant segment model yolov7-seg.pt in sahi

[kadirnar]

@pythoncodeAshisH
You should write the instance segmentation results in the bool_mask variable.
https://github.com/kadirnar/Yolov7-SAHI/blob/main/sahi/model.py#L1154

[kadirnar]

Updated SAHI version and added yolov7 pip support.

Yolov7Pip(Class): https://github.com/kadirnar/Yolov7-SAHI/blob/main/sahi/models/yolov7pip.py
Yolov7Hub(Class): https://github.com/kadirnar/Yolov7-SAHI/blob/main/sahi/models/yolov7hub.py

Demo_Code: Yolov7_Demo: https://github.com/kadirnar/Yolov7-SAHI/blob/main/demo/yolov7_demo.py
Result:

[msdiki]

Hi, getting the same error using the older version of this repo. When I use updated version with yolov7hub or yolov7pip, it automatically downloads a traced.pt and does not load my custom model. Earlier, it was working fine. Not sure what to try next.

result = get_sliced_prediction(
File "/home/sdki/.conda/envs/SAHI/lib/python3.10/site-packages/sahi-0.10.7-py3.10.egg/sahi/predict.py", line 239, in get_sliced_prediction
File "/home/sdki/.conda/envs/SAHI/lib/python3.10/site-packages/sahi-0.10.7-py3.10.egg/sahi/predict.py", line 91, in get_prediction
File "/home/sdki/.conda/envs/SAHI/lib/python3.10/site-packages/sahi-0.10.7-py3.10.egg/sahi/model.py", line 1114, in perform_inference
ValueError: Model is not loaded, load it by calling .load_model()

[danial880]

I am getting the same error with the new version. It is unable to load a custom model.

[kadirnar]

@danial880 , @msdiki

Can you share the sahi and yolov7 versions?

[danial880]

sahi: 0.11.6
yolov7detect: 0.3.1
@kadirnar, please let me know if you need something else

[kadirnar]

Where are you running the demo.py file? You have to run it in the yolov7-sahi/ directory. It may throw an error in the demo folder.

sahi: 0.11.6 yolov7detect: 0.3.1 @kadirnar, please let me know if you need something else

[danial880]

I am not using it inside sahi/directory. I have installed your repo using "python setup.py install", and used it in a script.

[kadirnar]

Orijinal Repo: https://github.com/obss/sahi
I just added yolov7 and tensorflowhub model support. To use this repo, you must:

git clone https://github.com/kadirnar/Yolov7-SAHI
cd Yolov7-SAHI

Open a new python file and run the following codes:

Yolov7: https://github.com/kadirnar/Yolov7-SAHI/blob/main/demo/yolov7_demo.py
TensorFlowHub: https://github.com/kadirnar/Yolov7-SAHI/blob/main/demo/tensorflowhub_demo.py

You have to run these files in the yolov7-sahi folder.

[CesareDavidePace]

how resolve this problem?

[kadirnar]

How did you install? What is the error message? @CesareDavidePace

[CesareDavidePace]

I just clone the repository.
And this is the error

[kadirnar]

Did you run the example yolov7-demo.py file inside the Yolov7-SAHİ folder?

[danial880]

I was facing a similar problem. In my scenario, I had to use yolov7-sahi in my project. Therefore I cannot use yolov7-demo.py from inside the yolov7-sahi folder. But the previous version of this repository works fine with custom models and runs from outside the yolov7-sahi folder after installation. Old Version

[kadirnar]

@danial880 @CesareDavidePace ,
The yolov7_hub module is running slow. You have to use the pip version. Can you download the repo and run the yolo7_demo.py file? I tested it now and it works.

git clone https://github.com/kadirnar/Yolov7-SAHI
cd Yolov7-SAHI

Env:

yolov7detect==0.3.1

Run

python yolov7_demo.py

Result:

[kadirnar]

This week I will add the SAHI algorithm to this library. You will use it more easily.
kadirnar/torchyolo#9

[CesareDavidePace]

[kadirnar]

pip uninstall sahi

[CesareDavidePace]

[kadirnar]

https://github.com/kadirnar/Yolov7-SAHI/blob/main/yolov7_demo.py

You should use this code.

[CesareDavidePace]

[kadirnar]

Did the yolov7.pt file download while running this file? Can you print for the self.model_path variable on line 20 of the Yolov7pip.py file?

[CesareDavidePace]

Yes, the file has been downloaded.
After install yolov7detect the script work.

But in my file i have the same problem

[CesareDavidePace]

Okay I restarted the kernel and everything works fine now. Thanks for your huge help!

[kadirnar]

This is great. This week I will release the pip package for the yolov7-sahi repository. Good luck.