June 2019
Intermediate to advanced
308 pages
7h 21m
English
We have trained and validated the MobileNet v1 model on the reduced acne dataset. The following screenshot presents the evaluation results of the model. As we can see from the screenshot, training accuracy is 1.0, or 100% in most steps, and final test accuracy is 0.89, or 89%:

The following graph, generated in TensorBoard using the model's training and validation accuracies' log files, show that training accuracies are a little inconsistent in the early steps, and consistently close to 1.0, or 100%, after that. On the other hand, validation accuracies are a little inconsistent in the range of 80-89%:
We believe ...
Read now
Unlock full access