Lec5: Building a Convolutional Neural Network from Scratch in Python for Classification of Images
In this video I have discussed Building a Convolutional Neural Network from Scratch in Python for Classification of Images. I have also explained the implementation of CNN models on Python. Following topics are discussed:
1:05 Acquisition of images. What is Pixels? Formation of Images.
10:48 System Design for the classification of Images.
11:46 Classical Approach for image classification.
12:27 Role of Feature Extraction in signals and Images.
21:49 Role of Domain Expert for selecting Feature Extraction method.
23:15 Standardization of Data: Min-Max Normalization and Z-Score Normalization.
24:12 Latent Representation of Data to reduce the dimensionality of the data.
25:10 Principal Component Analysis (PCA) to reduce the dimensionality of the data.
27:01 Feature Selection: Removing irrelevant features.
30:01 Feature Selection: Removing redundant features.
31:10 Preparing Dataset for the Classification: Training and Testing Dataset.
35:47 Artificial Neural Network (ANN).
45:36 Difference between Classical Approach and CNN Model to classify the images.
46:50 Convolutional Neural Network (CNN) Model to classify the images.
48:10 Convolution in CNN Model. Impact of applying different filters on the image.
56:17 Concept of ReLu Function.
57:58 Pooling Layer in CNN: Max Pooling and Average Pooling.
1:00:09 Flattening.
1:00:55 Fully Connected Layer.
1:01:30 Implementation of CNN Model in Python.
1:03:50 Loading Dataset and Displaying the sample data. Getting the number of samples in each class of the dataset.
1:11:11 Preprocessing of dataset and splitting the dataset.
1:12:26 Designing a CNN Model in Python.
1:23:27 Summarizing CNN Model and explanation regarding the number of parameters in each layer.
1:30:05 Compiling CNN Model: Selection of Optimizer, Loss Function and Metric.
1:33:10 Training of CNN Model: Concept of Batch size and Epoch, Displaying Loss and Accuracy with respect to Epoch.
1:35:43 CNN Model Evaluation using Testing Dataset: Confusion Matrix.
1:39:40 Displaying intermediate images obtained after each layer of CNN Model.