Deep Learning for Computer Vision

  1. Core Computer Vision Tasks
    1. Image Classification
      1. Problem Formulation
        1. Single-label Classification
          1. Multi-label Classification
            1. Hierarchical Classification
            2. Network Architecture Design
              1. Feature Extraction Backbone
                1. Classification Head
                  1. Output Layer Configuration
                  2. Training Strategies
                    1. Loss Function Selection
                      1. Optimization Techniques
                        1. Regularization Methods
                        2. Evaluation Metrics
                          1. Accuracy
                            1. Top-k Accuracy
                              1. Precision and Recall
                                1. F1 Score
                                  1. Area Under Curve
                                    1. Confusion Matrix Analysis
                                    2. Class Imbalance Handling
                                      1. Weighted Loss Functions
                                        1. Sampling Strategies
                                          1. Cost-sensitive Learning
                                        2. Object Detection
                                          1. Problem Definition
                                            1. Localization and Classification
                                              1. Multiple Object Handling
                                              2. Bounding Box Representation
                                                1. Coordinate Systems
                                                  1. Box Parameterization
                                                  2. Two-stage Detectors
                                                    1. R-CNN
                                                      1. Region Proposal
                                                        1. Feature Extraction
                                                          1. Classification and Regression
                                                          2. Fast R-CNN
                                                            1. ROI Pooling
                                                              1. Multi-task Loss
                                                                1. End-to-end Training
                                                                2. Faster R-CNN
                                                                  1. Region Proposal Network
                                                                    1. Anchor Generation
                                                                      1. Non-maximum Suppression
                                                                    2. Single-stage Detectors
                                                                      1. YOLO Family
                                                                        1. YOLOv1
                                                                          1. Grid-based Prediction
                                                                            1. Bounding Box Regression
                                                                            2. YOLOv2
                                                                              1. Anchor Boxes
                                                                                1. Batch Normalization
                                                                                2. YOLOv3
                                                                                  1. Multi-scale Prediction
                                                                                    1. Feature Pyramid
                                                                                    2. YOLOv4
                                                                                      1. Bag of Freebies
                                                                                        1. Bag of Specials
                                                                                        2. YOLOv5
                                                                                          1. Implementation Improvements
                                                                                        3. SSD
                                                                                          1. Multi-scale Feature Maps
                                                                                            1. Default Boxes
                                                                                              1. Hard Negative Mining
                                                                                              2. RetinaNet
                                                                                                1. Feature Pyramid Network
                                                                                                  1. Focal Loss
                                                                                                2. Anchor-based vs Anchor-free
                                                                                                  1. Anchor Design Challenges
                                                                                                    1. CenterNet
                                                                                                      1. FCOS
                                                                                                      2. Evaluation Metrics
                                                                                                        1. Intersection over Union
                                                                                                          1. Average Precision
                                                                                                            1. Mean Average Precision
                                                                                                              1. COCO Evaluation Protocol
                                                                                                            2. Semantic Segmentation
                                                                                                              1. Problem Definition
                                                                                                                1. Pixel-wise Classification
                                                                                                                  1. Dense Prediction
                                                                                                                  2. Fully Convolutional Networks
                                                                                                                    1. Removing Fully Connected Layers
                                                                                                                      1. Upsampling Strategies
                                                                                                                        1. Skip Connections
                                                                                                                        2. Encoder-Decoder Architectures
                                                                                                                          1. U-Net
                                                                                                                            1. Symmetric Architecture
                                                                                                                              1. Skip Connections
                                                                                                                                1. Medical Image Applications
                                                                                                                                2. SegNet
                                                                                                                                  1. Pooling Indices
                                                                                                                                    1. Memory Efficiency
                                                                                                                                    2. DeepLab Family
                                                                                                                                      1. Atrous Convolutions
                                                                                                                                        1. Atrous Spatial Pyramid Pooling
                                                                                                                                          1. CRF Post-processing
                                                                                                                                        2. Multi-scale Processing
                                                                                                                                          1. Pyramid Pooling Module
                                                                                                                                            1. Feature Pyramid Networks
                                                                                                                                            2. Evaluation Metrics
                                                                                                                                              1. Pixel Accuracy
                                                                                                                                                1. Mean Intersection over Union
                                                                                                                                                  1. Frequency Weighted IoU
                                                                                                                                                    1. Dice Coefficient
                                                                                                                                                  2. Instance Segmentation
                                                                                                                                                    1. Problem Definition
                                                                                                                                                      1. Object Detection and Segmentation
                                                                                                                                                        1. Instance Differentiation
                                                                                                                                                        2. Mask R-CNN
                                                                                                                                                          1. ROI Align
                                                                                                                                                            1. Mask Prediction Branch
                                                                                                                                                              1. Multi-task Loss
                                                                                                                                                              2. PANet
                                                                                                                                                                1. Path Aggregation Network
                                                                                                                                                                  1. Adaptive Feature Pooling
                                                                                                                                                                  2. YOLACT
                                                                                                                                                                    1. Real-time Instance Segmentation
                                                                                                                                                                      1. Prototype Generation
                                                                                                                                                                      2. Evaluation Metrics
                                                                                                                                                                        1. Mask Average Precision
                                                                                                                                                                          1. Boundary F1 Score
                                                                                                                                                                        2. Panoptic Segmentation
                                                                                                                                                                          1. Problem Definition
                                                                                                                                                                            1. Stuff and Things
                                                                                                                                                                              1. Unified Segmentation
                                                                                                                                                                              2. Panoptic FPN
                                                                                                                                                                                1. Semantic and Instance Branches
                                                                                                                                                                                2. UPSNet
                                                                                                                                                                                  1. Unified Architecture
                                                                                                                                                                                  2. Evaluation Metrics
                                                                                                                                                                                    1. Panoptic Quality

                                                                                                                                                                                Previous

                                                                                                                                                                                5. Modern CNN Architectures

                                                                                                                                                                                Go to top

                                                                                                                                                                                Next

                                                                                                                                                                                7. Advanced Topics and Applications