I trained a model but I don't know how to use it

General Discussion
#1

Hey guys, I’m a beginner in the ML area. I do have 2 years of experience in Python and C# but I just started to learn TensorFlow.

I inspired by Sebastian Lague’s Neural Network video and started the train a model to recognize 15 kinds of different doodles.

Even tho I managed to train a model with %90 accuracy idk how to use it. I wanna import 64x64 user generated images of doodles and feed it to the AI.

Here is the final code for the mentioned model:
tensorflow-quickdraw-recognition/doodle_identification_final_model.ipynb at main · beratcmn/tensorflow-quickdraw-recognition (github.com)

#3

Hi @beratcmn, you want to train the model on user generated images or you want to make predictions? Thank You.

#4

I was trying to predict user generated images and I figured it out. I’ll link the notebook, maybe someone can benefit from it.

There is a section at the bottom called “Uploading a custom file”
tensorflow-quickdraw-recognition/doodle_identification_final_model.ipynb at main · beratcmn/tensorflow-quickdraw-recognition · GitHub

#6

Make sure that the input data is the same format as the data that you used to train the model. if you are using a large model you may need to use a GPU to make predictions. you can also use load_weights() method on the model to load the weight of model from a file.