Transfer Learning script that operates on an SSD MobileNet V1 and V2 model pre-trained on a COCO image dataset
It lets users perform a transfer learning process on its own dataset of images
The training process can be watched with the Tensorboard visualization tool
Images have been labelled using graphical image annotation tool - LabelImg
Example result of a trained model that used a private dataset of Rubik's cube:
The newly trained model can be used with the ObjectToSpeech application to perform real-time detection https://github.com/tomsoch/ObjectToSpeech