Model json file parsing error

[yoo-jimin127]

Hello!
I found an error in parsing the /web_model/model.json file.
I referred to Issue #9 and #7 to find a solution, but the error was not resolved.
May I get some advice?

[zldrobit]

@yoo-jimin127 Of course. Plz share more information to debug the error:

• Have you changed the code in the repo?
• Did you link yolov5s_web_model to web_model, i.e. ln -s ../../yolov5/yolov5s_web_model public/web_model?
• Plz provide the output message from terminal after the command npm start?
• Press F12 in the browser to show the debug window and share the full debug stack.
• Run curl http://localhost:3000/web_model/model.json and curl http://localhost:3000/tfjs-yolov5-example/web_model/model.json on terminal, and share the output messages in terminal.

[yoo-jimin127]

Thank you very much for providing a detailed solution :)
Thanks to your advice, I was able to solve the problem.

I don't know exactly why, but ln -s../yolov5/yolov5s_web_modelpublic/web_model didn't work in my local environment.
Therefore, the problem was solved by replacing it with cp-r ../yolov5/yolov5s_web_model public/web_model.

Additionally, I am looking for a way to implement the function of changing the parameters of the yolov5 model in tf.js.
At https://github.com/ultralytics/yolov5,
It says that you can enter the yaml file and batch size in the following way for training.

python train.py --data coco.yaml --cfg yolov5n.yaml --weights '' --batch-size 128
                                       yolov5s                                64
                                       yolov5m                                40
                                       yolov5l                                24
                                       yolov5x                                16

In this code, Is there a way to train the model with the entered parameter value?

[zldrobit]

@yoo-jimin127 Yes, you could train whatever yolov5 model you like, e.g. yolov5s/m/l/x. After training, you could export the yolov5s/m/l/x model to tfjs format. Then, cp/ln -s the exported model path to the web_model path, e.g. tfjs-yolov5-example/public/web_model. Finally, run npm start as usual. Beware that yolov5m/l/x is slower than yolov5s on the browser.

[yoo-jimin127]

@zldrobit Thank you very much.
This repo is really useful for my project development.
Thank you for providing good data :)