Computer Vision
Object detection, segmentation, and backbone architectures — how deep learning models see and understand images.
6 lessons
01Object Detection
Anchor boxes, region proposals, YOLO, Faster R-CNN, FPN, and NMS — the full pipeline for finding and localising objects in images.
Segmentation
Semantic, instance, and panoptic segmentation — from fully convolutional networks to U-Net and Mask R-CNN.
Backbones in Practice
ResNet, MobileNet, EfficientNet, and transfer learning — choosing and fine-tuning the right backbone for detection and segmentation.
Vision-Language Models
CLIP's contrastive image-text training, the shared embedding space, zero-shot classification, and the BLIP-2 / LLaVA recipe for plugging vision into an LLM.
Self-Supervised Learning
Learning useful representations without labels — pretext tasks, contrastive methods (SimCLR, MoCo), masked image modeling (MAE), and why self-supervision is what made foundation models possible.
Computer Vision Quiz
Test your understanding of object detection, segmentation, backbones, vision-language models, and self-supervised learning.