Build a Convolutional Neural Network (CNN) using PyTorch - Example with 'USPS' dataset
In this video, I tried to perform image classification by implementing Convolutional Neural Network (CNN). Convolutional neural networks use convolution operation instead of linear layers like artificial neural networks. I used the USPS dataset, which contains handwritten digit images of 16 x 16 pixels. CNNs include many layers such as convolutional layers, non-linear activation layers, pooling layers and fully connected layers. These types of deep learning frameworks are useful for image classification tasks.
GitHub address: https://github.com/randomaccess2023/M...
For more details: https://en.wikipedia.org/wiki/Convolu...
Starting point of each section:
01:19 --- Import the libraries
02:36 --- Define the device
05:01 --- Load USPS dataset
10:31 --- Define dataloaders
12:07 --- Define the network architecture
18:40 --- Visualize the model architecture
20:50 --- Define the train batch
22:23 --- Define the test batch
23:48 --- Define the accuracy function
26:34 --- Set up model (CNN), criterion (loss function) and optimizer
27:30 --- Train the model
35:52 --- Plot model loss and accuracy
41:11 --- Model prediction
#neural_network #supervised_learning #image_classification #usps_dataset #data_science #jupyter_notebook #python #pytorch #cnn