Series

Reads and papers written to be read in order. Each series has its own page with the highlights of every part, so you can see where it goes before you start.

RT-DETR

You can explain how each RT-DETR version makes DETRs faster and stronger: v1 beats YOLOs with a hybrid encoder and query selection, v2 adds decoder dials and training freebies, v3 adds training-only positives, v4 distills DINOv3 features, and World enables open-vocabulary detection.

5 parts · 17 interactive figures

  1. 1RT-DETR — DETRs Beat YOLOs, Encoder by Encoder
  2. 2RT-DETRv2 — A Bag of Freebies for the Decoder
  3. 3RT-DETRv3 — More Positives Without Paying at Inference
  4. 4RT-DETRv4 — Borrowing a Foundation Model's Eyes
  5. 5RT-DETR-World — Naming Any Object in Real Time
Open the series →

Multiple View Geometry

You can calibrate a camera, correct its lens, recover its pose from known points or from matches to a second view, and turn two images into a metric point cloud by rectifying, matching, and triangulating every pixel.

11 parts · 29 interactive figures

  1. 1The Pinhole Camera — A Pixel Is a Ray
  2. 2Where Is the Camera? — R, C and t
  3. 3Points at Infinity — Horizons, Vanishing Points and the Floor
  4. 4Finding the Camera From Known Points — P by Least Squares
  5. 5Lens Distortion — Bending Straight Lines Back
  6. 6What calibrateCamera Actually Does
  7. 7Epipolar Geometry — A Point Becomes a Line
  8. 8Triangulation — Two Rays That Almost Meet
  9. 9From Matches to Motion — Four Cameras, One Right Answer
  10. 10Stereo Depth — Rectification and Z = fB/d
  11. 11From Two Photos to a Point Cloud
Open the series →

Transformers in Vision

After this series you will be able to read any vision paper, count the FLOPs of any attention layer, write the losses of DETR and CLIP from memory, and explain why convolutions and transformers win or lose.

12 parts · 24 interactive figures

  1. 1Attention Before Transformers — Score, Squash, Multiply
  2. 2Query, Key, Value — Dot-Product Attention from Scratch
  3. 3Attention Is All You Need — The Transformer, Block by Block
  4. 4Attention Meets Pixels — Windows, Axes and the First Fully Attentional Networks
  5. 5DETR — Detection as Set Prediction, No Anchors and No NMS
  6. 6An Image Is Worth 16×16 Words — ViT, Patch by Patch
  7. 7DeiT — Training a Vision Transformer on ImageNet Alone
  8. 8Swin Transformer — Windows, Shifts and a Pyramid for Dense Prediction
  9. 9Convolutions Strike Back — Was It the Attention or the Recipe?
  10. 10Masked Autoencoders — Hide 75 %, Encode the Rest
  11. 11CLIP — Learning Pictures From the Words Next to Them
  12. 12Vision Transformers Today — May 2022 to October 2026, Year by Year
Open the series →

YOLOv4

Understand YOLOv4's architecture, its Bag of Freebies and Specials, and why it is a fast, reliable real-time object detector.

5 parts · 6 interactive figures

  1. 1YOLOv4 — Version 0: Introduction
  2. 2YOLOv4 — Version 1: Bag of Freebies
  3. 3YOLOv4 — Version 2: Bag of Specials
  4. 4YOLOv4 — Version 3: Proposed Workflow
  5. 5YOLOv4 — Version 4: Final Verdict
Open the series →

Text2SQL

5 parts · by Devshree Patel on VisionWizard

  1. 1Text2SQL — Part 1: Introduction
  2. 2Text2SQL — Part 2: Datasets
  3. 3Text2SQL — Part 3: Baseline Models
  4. 4Text2SQL — Part 4: State-Of-The-Art Models
  5. 5Text2SQL — Part 5: Final Insights
Open the series →