Computer Vision
From pixels to perception: image processing, CNNs, detection, segmentation, pose, tracking, OCR and vision transformers.
What Is Computer Vision?
The field, the tasks, and the datasets that drive progress in visual intelligence.
Image Representation
Pixels, channels, color spaces and the tensor layouts every vision library uses.
Image Processing Fundamentals
Convolutions, blur, edges and thresholds — the classic toolbox that still powers pipelines today.
Image Augmentation
Synthesize training variety: flips, rotations, crops and color jitter to make models robust.
Image Classification
The canonical vision task: train a classifier on MNIST/CIFAR and read the results.
CNNs for Vision
Architecture patterns that win on images: receptive fields, pooling, and depth.
Transfer Learning for Vision
Pretrained backbones are the default starting point — swap the head, keep the features.
Object Detection
Bounding boxes and classes: two-stage detectors, single-shot detectors, and YOLO in practice.
Semantic Segmentation
Label every pixel: FCNs, U-Net, and the encoder-decoder design for pixel-level tasks.
Instance Segmentation
Separate each object, not just each class: Mask R-CNN and the mask-head idea.
Pose Estimation
Detect body keypoints and skeletons with heatmap-based and regression approaches.
OpenCV Fundamentals
The industrial workhorse: reading, writing, resizing, drawing, and real-time camera capture.
Feature Detection & Matching
Keypoints and descriptors: SIFT, ORB, and how image matching enables stitching and localization.
Image Registration & Stitching
Combine overlapping images into panoramas and align images across time or sensors.
Video Analysis & Tracking
Optical flow, frame differencing, and object tracking across video streams.
Face Detection & Recognition
Detect faces, embed them into vectors, and recognize identity — and respect the ethics of it.
OCR & Document AI
Turn images of text into machine-readable text: Tesseract, PaddleOCR, and layout understanding.
3D Vision
Depth, point clouds and stereo: how cameras recover the third dimension.
Vision Transformers (ViT)
Patch the image into tokens and let attention do the rest — the modern alternative to CNNs.
Evaluating Vision Models
Benchmarks, metrics and failure analysis: accuracy, mAP, and per-class audits.
Computer Vision Roadmap
Synthesize the course into a plan: pick a specialization, build portfolio projects, and keep shipping.