-
Notifications
You must be signed in to change notification settings - Fork 0
Logs
Sean Trinh edited this page May 16, 2019
·
2 revisions
With this wiki page, I will be documenting and logging my progress in improving my model for this binary classification problem. Logs can be found below.
The model currently on the repository demonstrated a 72% test accuracy. I used the VGG16 network architecture and added it to my model, but set its parameters to be non-trainable. I am currently still facing issues of overfitting, since the data available is very limited, since training accuracy exceeds validation accuracy after epoch 3. The model needs to be able to generalize features, rather than become really good at classifying the given set of images in the training data set. My next goal is to improve the model's accuracy by first reducing overfitting.