Mastering Computer Vision with TensorFlow 2.x

Iteration 1 – CNN parameter adjustment

Drop one Conv2D 64 and one Conv2D 128 so that the CNN has only one Conv2D 64 and one Conv2D 128 with this change:

Epoch 1/5 256/256 [========================] - 63s 247ms/step - loss: 1.7497 - acc: 0.2805 Epoch 2/5 256/256 [========================] - 64s 248ms/step - loss: 1.5192 - acc: 0.4095 Epoch 3/5 256/256 [========================] - 65s 252ms/step - loss: 1.3553 -acc: 0.4832 Epoch 4/5 256/256 [========================] - 66s 260ms/step - loss: 1.2633 - acc: 0.5218 Epoch 5/5 256/256 [========================] - 65s 256ms/step - loss: 1.1919 - acc: 0.5483

Result: Dropping a Conv2D layer did not adversely affect the performance but did not make it better either.

Next, we are going to keep the change ...