Hello,
I have tried to play with the template code. I have written a data loader for NYU dataset. During debugging, I used mini version of dataset and also created a mini convolutional network. For now, the overall framework is able to load data, procedure the training, visualize loss(Tensorboard) and also make prediction for NYU dataset.
Some issues related to data loader;
I haven't used the data API of tensorflow, is it really necessary?
I have changed the batch selection little bit, please check and give feedback to me.
Also I have tried to open two notebooks for training and prediction but could not use to Google Colaboratory. If some babayigit(No offense :) ) is able to do it, I would be very happy.
Hello, I have tried to play with the template code. I have written a data loader for NYU dataset. During debugging, I used mini version of dataset and also created a mini convolutional network. For now, the overall framework is able to load data, procedure the training, visualize loss(Tensorboard) and also make prediction for NYU dataset.
Some issues related to data loader;
Also I have tried to open two notebooks for training and prediction but could not use to Google Colaboratory. If some babayigit(No offense :) ) is able to do it, I would be very happy.