Sign inStart free
AI Engineering/Phase 04
Core

Computer Vision

From convolutions to detection, segmentation and vision transformers.

28 lessons ~31.3 hr
  1. Image Fundamentals — Pixels, Channels, Color SpacesBuild · Free · 45 minAn image is a tensor of light samples. Every vision model you will ever use starts from this one fact.Open
  2. Convolutions from ScratchBuild · 1 h 15 minA convolution is a tiny dense layer you slide across an image, sharing the same weights at every location.Needs Full Access
  3. CNNs — LeNet to ResNetLearn + Build · 1 h 15 minEvery major CNN of the last thirty years is the same conv–nonlinearity–downsample recipe with one new idea bolted on. Learn the ideas in order.Needs Full Access
  4. Image ClassificationBuild · 1 h 15 minA classifier is a function from pixels to a probability distribution over classes. Everything else is plumbing.Needs Full Access
  5. Transfer Learning & Fine-TuningBuild · 1 h 15 minSomebody else spent a million GPU hours teaching a network what edges, textures, and object parts look like. You should borrow those features before training your own.Needs Full Access
  6. Object Detection — YOLO from ScratchBuild · 1 h 15 minDetection is classification plus regression, run at every position in a feature map, then cleaned up with non-maximum suppression.Needs Full Access
  7. Semantic Segmentation — U-NetBuild · 1 h 15 minSegmentation is classification at every pixel. U-Net makes it work by pairing a downsampling encoder with an upsampling decoder and wiring skip connections between them.Needs Full Access
  8. Instance Segmentation — Mask R-CNNBuild + Learn · 1 h 15 minAdd a tiny mask branch to a Faster R-CNN detector and you have instance segmentation. The hard part is RoIAlign, and it is harder than it looks.Needs Full Access
  9. Image Generation — GANsBuild · 1 h 15 minA GAN is two neural networks in a fixed game. One draws, one critiques. They get better together until the drawings fool the critic.Needs Full Access
  10. Image Generation — Diffusion ModelsBuild · 1 h 15 minA diffusion model learns to denoise. Train it to remove a tiny bit of noise from a noisy image, repeat that backwards a thousand times, and you have an image generator.Needs Full Access
  11. Stable Diffusion — Architecture & Fine-TuningLearn + Use · 1 h 15 minStable Diffusion is a DDPM that runs in the latent space of a pretrained VAE, conditioned on text via cross-attention, sampled with a fast deterministic ODE solver, and steered by classifier-free guidance.Needs Full Access
  12. Video Understanding — Temporal ModelingLearn + Build · 45 minA video is a sequence of images plus the physics that connects them. Every video model either treats time as an extra axis (3D conv), a sequence to attend over (transformer), or a feature to extract once and pool (2D+pool).Needs Full Access
  13. 3D Vision — Point Clouds & NeRFsLearn + Build · 45 min3D vision comes in two flavours. Point clouds are the sensor's raw output. NeRFs are the learned volumetric field. Both answer "what is where in space."Needs Full Access
  14. Vision Transformers (ViT)Build · 45 minCut the image into patches, treat each patch as a word, run a standard transformer. Don't look back.Needs Full Access
  15. Real-Time Vision — Edge DeploymentLearn + Build · 1 h 15 minEdge inference is the discipline of getting a 90-accuracy model to run at 30 fps on a device with 2 GB of RAM. Every percentage point of accuracy is traded against milliseconds of latency.Needs Full Access
  16. Build a Complete Vision Pipeline — CapstoneBuild · 2 hA production vision system is a chain of models and rules stitched with data contracts. The pieces are already in this phase; the capstone wires them together end-to-end.Needs Full Access
  17. Self-Supervised Vision — SimCLR, DINO, MAELearn + Build · 1 h 15 minLabels are the bottleneck of supervised vision. Self-supervised pretraining removes them: learn visual features from 100M unlabelled images, fine-tune on 10k labelled ones.Needs Full Access
  18. Open-Vocabulary Vision — CLIPBuild + Use · 45 minTrain an image encoder and a text encoder together so that matching (image, caption) pairs land at the same point in a shared space. That is the whole trick.Needs Full Access
  19. OCR & Document UnderstandingLearn + Use · 45 minOCR is a three-stage pipeline — detect text boxes, recognise the characters, then lay them out. Every modern OCR system reorders these stages or merges them.Needs Full Access
  20. Image Retrieval & Metric LearningBuild · 45 minA retrieval system ranks candidates by a distance in embedding space. Metric learning is the discipline of shaping that space so the distances mean what you want.Needs Full Access
  21. Keypoint Detection & Pose EstimationBuild · 45 minA pose is a set of ordered keypoints. A keypoint detector is a heatmap regressor. Everything else is bookkeeping.Needs Full Access
  22. 3D Gaussian Splatting from ScratchBuild · 1 h 30 minA scene is a cloud of millions of 3D Gaussians. Each one has a position, orientation, scale, opacity, and a colour that depends on viewing direction. Rasterise them, backprop through the rasterisation, done.Needs Full Access
  23. Diffusion Transformers & Rectified FlowLearn + Build · 1 h 15 minThe U-Net is not the secret of diffusion. Replace it with a transformer, swap the noise schedule for a straight-line flow, and suddenly you have SD3, FLUX, and every 2026 text-to-image model.Needs Full Access
  24. SAM 3 & Open-Vocabulary SegmentationUse + Build · 1 hGive a model a text prompt and an image and get masks for every matching object. SAM 3 made that a single forward pass.Needs Full Access
  25. Vision-Language Models — The ViT-MLP-LLM PatternLearn + Use · 1 h 15 minA vision encoder converts an image into tokens. An MLP projector maps those tokens into the LLM's embedding space. A language model does the rest. That pattern — ViT-MLP-LLM — is every production VLM in 2026.Needs Full Access
  26. Monocular Depth & Geometry EstimationBuild + Use · 1 hA depth map is a single-channel image where each pixel is a distance from the camera. Predicting it from one RGB frame used to be impossible without stereo or LiDAR. In 2026 a frozen ViT encoder plus a lightweight head gets within a few percent of ground truth.Needs Full Access
  27. Multi-Object Tracking & Video MemoryBuild · 1 hTracking is detection plus association. Detect every frame. Match this frame's detections to last frame's tracks by ID.Needs Full Access
  28. World Models & Video DiffusionLearn + Build · 1 h 15 minA video model that predicts the next seconds of a scene is a world simulator. Condition that prediction on actions and you have a learned game engine.Needs Full Access