Deep Learning for Images | Incognition

Deep Learning for Images

Deep Learning for Images

Why ordinary layers fail on pixels, and what convolution does instead — from a grid of numbers to the architectures that made computer vision work.

0% completed

0 done40 lessons

A Picture Is a Grid of Numbers

Colour Channels

Why Flattening an Image Loses Meaning

Spatial Structure

Edges and Simple Patterns

Filters as Pattern Detectors

Sliding a Filter

Image Preprocessing

The Vision Problem, Framed

The Convolution Operation

Multiple Filters

Max vs Average Pooling

Stacking Convolutional Layers

The Receptive Field

Parameter Sharing

Anatomy of a CNN

From Features to a Prediction

LeNet and the First CNNs

Residual Connections

Inception and Parallel Paths

Comparing Architectures

What a CNN Actually Learns

Transfer Learning

Freezing and Fine-Tuning

Working With Small Datasets

Image Classification End to End

Object Detection

Bounding Boxes and IoU

Semantic Segmentation

Data Augmentation for Vision

Common Vision Failure Modes

Vision Beyond CNNs