Where can I find the training dataset?

[zhewang95]

or if the dataset will not be published, can you tell me how many images are used in training?

[soufianesabiri]

I want to retrain the model for other types of license plates, but I don't find any way on how to do it...

[panda-lab]

@zhewang95 about 8000 images

[panda-lab]

@soufianesabiri you can refer this blog https://blog.csdn.net/weixin_39750664/article/details/82502302

[soufianesabiri]

@lsy17096535 Thanks! I found this also: https://tolotra.com/2018/09/15/how-to-retrain-ssd-mobilenet-for-real-time-object-detection-using-a-raspberry-pi-and-movidius-neural-compute-stick/

Can I use it? Is it the same procedure?

Also, there is this file: https://github.com/tensorflow/hub/blob/master/examples/image_retraining/retrain.py

Which I found in another article, they used a command like this:

python retrain.py --how_many_training_steps=500 --output_graph=out.pb --output_labels=labels.txt --image_dir=lp --tfhub_module https://tfhub.dev/google/imagenet/mobilenet_v1_100_224/feature_vector/1

To generate the pb file mentionned in the HyperLPR baidu link: https://pan.baidu.com/s/11V0Pfx6jGNaBOOTqDSVK2Q

[panda-lab]

@soufianesabiri If you want to train a caffe model, you can refer this blog. Other framework you should read similar blog.

[soufianesabiri]

Ok I will, thank you for your answer!