Convolutional Neural Networks

Course 4 of 5 in the Deep Learning Specialization by Andrew Ng (DeepLearning.AI).

Topics

Course 4 of 5 in the Deep Learning Specialization by Andrew Ng (DeepLearning.AI).

Notes

Convolutions and Edge Detection
deep-learning
convolutional-neural-networks
computer-vision
edge-detection
padding

Why large images break fully connected networks, how the convolution operation detects vertical and horizontal edges, and how padding preserves image size.

Aug 11, 2026

Strided Convolutions and Convolutions Over Volume
deep-learning
convolutional-neural-networks
computer-vision
convolution
stride

How a stride changes the output size of a convolution, why deep learning skips the filter flip, and how filters extend to RGB volumes and to many filters at once.

Aug 13, 2026

One Layer of a Convolutional Network
deep-learning
convolutional-neural-networks
computer-vision
convolution
notation

How filters, a bias, and a non-linearity form one convolutional layer, why the parameter count does not depend on image size, and the notation for a deep ConvNet.

Aug 13, 2026

Pooling Layers
deep-learning
convolutional-neural-networks
computer-vision
pooling

How max pooling shrinks a representation and keeps the strongest feature responses, how average pooling differs, and why pooling has hyperparameters but no parameters.

Aug 13, 2026

Full ConvNet Example and Why Convolutions
deep-learning
convolutional-neural-networks
computer-vision
architecture
parameter-sharing

A complete LeNet-inspired ConvNet with its activation shapes and parameter counts, and why parameter sharing and sparse connections make convolutions work so well.

Aug 13, 2026

Lab: Convolutional Model Step by Step
deep-learning
convolutional-neural-networks
computer-vision
numpy
lab

Implement zero padding, a single convolution step, and the forward and backward passes of the convolution and pooling layers in pure NumPy.

Aug 14, 2026

Lab: Convolutional Model Application
deep-learning
convolutional-neural-networks
computer-vision
tensorflow
keras
lab

Build and train two ConvNets in TensorFlow Keras, a smile detector with the Sequential API and a sign language classifier with the Functional API.

Aug 14, 2026

Classic Networks
deep-learning
convolutional-neural-networks
computer-vision
architecture
lenet
alexnet
vgg

LeNet-5, AlexNet, and VGG-16, the three architectures that established the patterns every later convolutional network is built on.

Aug 14, 2026

Modern Networks
deep-learning
convolutional-neural-networks
computer-vision
architecture
resnet
skip-connections
network-in-network
inception
googlenet

Skip connections and ResNets, why a one by one filter does more than multiply, and how bottlenecks make the inception module affordable enough to stack.

Aug 15, 2026

Lab: Residual Networks
deep-learning
convolutional-neural-networks
computer-vision
resnet
skip-connections
tensorflow
keras
lab

Build the identity block and the convolutional block in Keras, assemble them into a 50 layer ResNet, and train it to recognize sign language digits.

Aug 15, 2026

MobileNet and EfficientNet
deep-learning
convolutional-neural-networks
computer-vision
architecture
mobilenet
efficientnet

How a depthwise separable convolution cuts cost by roughly ten times, how MobileNet v2 adds expansion and projection, and how EfficientNet scales a network to a device.

Aug 15, 2026

Practical Advice for Using ConvNets
deep-learning
convolutional-neural-networks
computer-vision
transfer-learning
data-augmentation

Starting from open source implementations, transferring pretrained weights to a small dataset, augmenting images, and what the state of computer vision implies.

Aug 16, 2026

Lab: Transfer Learning with MobileNetV2
deep-learning
convolutional-neural-networks
computer-vision
transfer-learning
mobilenet
fine-tuning
data-augmentation
tensorflow
keras
lab

Turn an ImageNet MobileNetV2 into an alpaca detector by training one new neuron on frozen features, then fine tuning the last 34 layers of the network.

Aug 17, 2026

Object Localization and Landmark Detection
deep-learning
convolutional-neural-networks
computer-vision
object-detection
localization
landmarks

How a ConvNet learns to output bounding box coordinates alongside a class label, the eight component target vector, and how the same idea locates facial landmarks.

Aug 17, 2026

Sliding Windows and Region Proposals
deep-learning
convolutional-neural-networks
computer-vision
object-detection
sliding-windows
r-cnn

Why classifying every window in an image is too slow, how turning fully connected layers into convolutions collapses the sweep into one forward pass, and what region…

Aug 17, 2026

Bounding Box Predictions with YOLO
deep-learning
convolutional-neural-networks
computer-vision
object-detection
yolo

How placing a grid on the image and assigning each object to the cell holding its midpoint turns detection into one convolutional pass with precise boxes.

Aug 18, 2026

IoU, Non-max Suppression, and Anchor Boxes
deep-learning
convolutional-neural-networks
computer-vision
object-detection
yolo
non-max-suppression
anchor-boxes

How overlap is measured, how duplicate detections of the same object are removed, how one grid cell reports two objects, and how the whole YOLO algorithm fits together.

Aug 18, 2026

Lab: Car Detection with YOLO
deep-learning
convolutional-neural-networks
computer-vision
object-detection
yolo
tensorflow
non-max-suppression
lab

Use a pretrained YOLOv2 detector on a road image, filter its boxes, apply class-aware non-max suppression, and draw final detections.

Aug 18, 2026

Semantic Segmentation with U-Net
deep-learning
convolutional-neural-networks
computer-vision
semantic-segmentation
u-net
transpose-convolution

Labeling every pixel rather than drawing a box, the transpose convolution that grows a representation back, and the U-Net that puts the two together.

Aug 18, 2026

Lab: Image Segmentation with U-Net
deep-learning
convolutional-neural-networks
computer-vision
semantic-segmentation
u-net
transpose-convolution
tensorflow
keras
lab

Build a U-Net in Keras from an encoder block and a decoder block, then train it to label every pixel of a driving scene with one of 23 classes.

Aug 19, 2026

Face Recognition and One Shot Learning
deep-learning
convolutional-neural-networks
computer-vision
face-recognition
one-shot-learning
verification

Why recognition is harder than verification, why a softmax cannot learn from one picture per person, and how a learned similarity function solves both.

Aug 20, 2026

Siamese Networks and the Triplet Loss
deep-learning
convolutional-neural-networks
computer-vision
face-recognition
siamese-network
triplet-loss
embeddings

Turning a picture into a 128 number encoding, the triplet loss and the margin that trains one, and face verification posed as binary classification.

Aug 20, 2026

Lab: Face Recognition
deep-learning
convolutional-neural-networks
computer-vision
face-recognition
siamese-network
triplet-loss
embeddings
tensorflow
keras
lab

Load a pretrained FaceNet, implement the triplet loss, encode a database of faces, and build verification and recognition on top of the distances.

Aug 20, 2026

Neural Style Transfer
deep-learning
convolutional-neural-networks
computer-vision
neural-style-transfer
visualization
gram-matrix
cost-function
convolution
signals

Repainting a photograph in the style of an artwork, the two costs that make it work, and what convolution looks like over 1D and 3D data.

Aug 21, 2026

Lab: Art Generation with Neural Style Transfer
deep-learning
convolutional-neural-networks
computer-vision
neural-style-transfer
gram-matrix
vgg
tensorflow
keras
lab

Write the content cost, the Gram matrix and the style cost in TensorFlow, then run gradient descent on the pixels of an image until a photograph of the Louvre comes back…

Aug 21, 2026
No matching items
Back to top