Deep Learning with PyTorch

  1. Convolutional Neural Networks (CNNs) for Computer Vision
    1. CNN Fundamentals
      1. Convolution Operation
        1. Mathematical Definition
          1. Discrete Convolution
            1. Cross-Correlation vs Convolution
            2. Local Receptive Fields
              1. Spatial Locality
                1. Feature Detection
                2. Parameter Sharing
                  1. Weight Sharing
                    1. Translation Invariance
                    2. Feature Maps
                      1. Channel Concept
                        1. Feature Hierarchy
                        2. Stride and Padding
                          1. Output Size Calculation
                            1. Valid vs Same Padding
                          2. CNN Architecture Design
                            1. Convolutional Blocks
                              1. Conv-ReLU-Pool Pattern
                                1. Residual Blocks
                                  1. Dense Blocks
                                  2. Feature Extraction Layers
                                    1. Early Layer Features
                                      1. Deep Layer Features
                                      2. Classification Head
                                        1. Global Average Pooling
                                          1. Fully Connected Layers
                                            1. Output Layer Design
                                          2. Advanced CNN Concepts
                                            1. Dilated Convolutions
                                              1. Atrous Convolutions
                                                1. Receptive Field Expansion
                                                2. Grouped Convolutions
                                                  1. Channel Grouping
                                                    1. Computational Efficiency
                                                    2. Depthwise Separable Convolutions
                                                      1. Depthwise Convolution
                                                        1. Pointwise Convolution
                                                          1. MobileNet Architecture
                                                        2. Transfer Learning
                                                          1. Pre-trained Models
                                                            1. ImageNet Pre-training
                                                              1. Model Zoo (torchvision.models)
                                                              2. Feature Extraction
                                                                1. Freezing Layers
                                                                  1. Feature Extractor Setup
                                                                  2. Fine-tuning
                                                                    1. Layer Unfreezing Strategies
                                                                      1. Learning Rate Scheduling
                                                                        1. Gradual Unfreezing
                                                                      2. Classic CNN Architectures
                                                                        1. LeNet-5
                                                                          1. Historical Significance
                                                                            1. Architecture Details
                                                                            2. AlexNet
                                                                              1. Deep Learning Breakthrough
                                                                                1. Key Innovations
                                                                                2. VGG Networks
                                                                                  1. VGG-16 and VGG-19
                                                                                    1. Deep and Uniform Architecture
                                                                                    2. ResNet
                                                                                      1. Residual Connections
                                                                                        1. Skip Connections
                                                                                          1. Deep Network Training
                                                                                          2. Inception Networks
                                                                                            1. Inception Modules
                                                                                              1. Multi-scale Feature Processing
                                                                                              2. DenseNet
                                                                                                1. Dense Connections
                                                                                                  1. Feature Reuse
                                                                                                2. Modern CNN Architectures
                                                                                                  1. EfficientNet
                                                                                                    1. Compound Scaling
                                                                                                      1. Neural Architecture Search
                                                                                                      2. Vision Transformer (ViT)
                                                                                                        1. Attention in Computer Vision
                                                                                                          1. Patch-based Processing

                                                                                                      Previous

                                                                                                      5. Training and Evaluating Models

                                                                                                      Go to top

                                                                                                      Next

                                                                                                      7. Recurrent Neural Networks (RNNs) for Sequential Data