8 steps · about 57 min · 2012–2023

Computer vision, AlexNet to SAM

How deep networks took over image understanding: the convolutions that started it, residual connections that made depth trainable, the transformer and contrastive turns, real-time detection, segmentation and neural scene representations.

Fig. 02 / Reading route8 papers, in order
Reading path: AlexNet: ImageNet Classification with Deep Convolutional Neural Networks; ResNet: Deep Residual Learning for Image Recognition; You Only Look Once: Unified, Real-Time Object Detection; U-Net: Convolutional Networks for Biomedical Image Segmentation; An Image is Worth 16×16 Words: Transformers for Image Recognition at Scale; CLIP: Learning Transferable Visual Models From Natural Language Supervision; NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis; Segment Anything0102030405060708
Start with AlexNet: ImageNet Classification with Deep Convolutional Neural Networks. Follow the numbered steps to Segment Anything.
  1. AlexNet: ImageNet Classification with Deep Convolutional Neural Networks 2012 · 7 min read · Beginner — AlexNet won ImageNet 2012 by a wide margin with GPU-trained deep convolutions, ReLU and dropout — the result that started the deep learning era.
  2. ResNet: Deep Residual Learning for Image Recognition 2015 · 7 min read · Beginner — An AI-generated review of ResNet, the 2015 architecture that used skip connections to train networks hundreds of layers deep and became the default backbone.
  3. You Only Look Once: Unified, Real-Time Object Detection 2015 · 7 min read · Beginner — An AI-generated review of YOLO, the 2015 detector that framed object detection as a single regression problem and made real-time detection possible on a GPU.
  4. U-Net: Convolutional Networks for Biomedical Image Segmentation 2015 · 7 min read · Beginner — U-Net's symmetric encoder-decoder with skip connections became the standard for pixel-level prediction, from biomedical imaging to diffusion models.
  5. An Image is Worth 16×16 Words: Transformers for Image Recognition at Scale 2020 · 7 min read · Intermediate — The Vision Transformer applied a plain transformer to image patches and matched CNNs at scale — proving convolutions were optional for recognition.
  6. CLIP: Learning Transferable Visual Models From Natural Language Supervision 2021 · 8 min read · Advanced — An AI-generated review of CLIP, the contrastive vision-language model that made zero-shot image classification and text-guided image understanding practical.
  7. NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis 2020 · 8 min read · Advanced — An AI-generated review of NeRF, the 2020 method that represents a 3D scene as a neural field and renders photorealistic novel views from sparse input photos.
  8. Segment Anything 2023 · 6 min read · Intermediate — How the Segment Anything project reframed segmentation as a promptable task and released a 1-billion-mask dataset that made zero-shot segmentation practical.

Every step as a full report

Prefer to jump around? Each step is a standalone report with its own citations, figures and FAQ.

ResNet: Deep Residual Learning for Image Recognition

Architectures · 7 min read · Beginner

An AI-generated review of ResNet, the 2015 architecture that used skip connections to train networks hundreds of layers deep and became the default backbone.

cnnresidual-connectionscomputer-vision Read

You Only Look Once: Unified, Real-Time Object Detection

Computer vision · 7 min read · Beginner

An AI-generated review of YOLO, the 2015 detector that framed object detection as a single regression problem and made real-time detection possible on a GPU.

yoloobject-detectioncomputer-vision Read

Segment Anything

Computer vision · 6 min read · Intermediate

How the Segment Anything project reframed segmentation as a promptable task and released a 1-billion-mask dataset that made zero-shot segmentation practical.

segmentationcomputer-visionfoundation-models Read