DL_Assignment3

NCKU 1112_深度學習 DEEP LEARNING

1. Implement LeNet5 (any version) by Tensorflow (Keras is ok) and Pytorch.

1.1 Tensorflow

https://www.tensorflow.org/overview?hl=zh-cn

1.1.1 Split dataset

Only do once.

1.1.2 Load dataset

1.1.3 Dynamic

1.2 Pytorch

https://github.com/gradient-ai/LeNet5-Tutorial?ref=blog.paperspace.com

2. Performance Comparison (accuracy, inference time, space complexity, and flops) of the TF/Pytorch and your handcrafted versions.

2.1 My handcrafted versions

2.2 Performance Comparision

3. In the TensorFlow version, please make a "static graph" LeNet5 and compare it with a "dynamic" one in inference and training time.

3.1 Static graph