> ## Documentation Index > Fetch the complete documentation index at: https://aegean.ai/llms.txt > Use this file to discover all available pages before exploring further. # Perception > Visual perception with CNNs, object detection, and segmentation. This chapter covers visual perception: convolutional networks, object detection, and pixel-level segmentation. Sensor modeling (cameras, lidar, calibration, beam models) lives in the [Robotics Systems](/book/robotics-systems/index) chapter, where it sits alongside state estimation and SLAM. Convolutional neural networks for image classification, including layer types, architectures, and visualization techniques. Detecting and localizing objects in images using two-stage and single-stage deep learning detectors. Pixel-wise classification for semantic and instance segmentation using Mask RCNN and UNet. ## Convolutional neural networks CNNs are the backbone of modern computer vision systems, enabling image classification, feature extraction, and visual understanding. Introduction to convolutional neural networks. Understanding CNN layer types and operations. Example architectures: LeNet, AlexNet, VGG, ResNet. ResNet as a backbone for downstream vision tasks. ## Object detection Object detection covers scene understanding fundamentals, evaluation metrics, and the evolution from two-stage (RCNN family) to single-stage (YOLO family) detectors. Detection vs classification, the detection pipeline, region proposals, FCNs, and the COCO dataset. Precision, recall, mAP, and IoU for evaluating detectors. Region-based CNN: selective search, CNN features, SVM classification. Shared convolutional features and ROI pooling for end-to-end training. Region Proposal Network enabling fully end-to-end two-stage detection. ### Faster RCNN from scratch (PyTorch) A six-notebook series building every Faster RCNN component from scratch in pure PyTorch, from COCO data loading through end-to-end training and inference. Streaming COCO from Hugging Face, collation, and anchor target assignment. ResNet50 feature pyramid network (FPN) with lateral connections. Region Proposal Network: anchor generation, objectness head, NMS. ROI Align, two-layer MLP head, and sibling classification and regression predictors. End-to-end training with AMP and gradient checkpointing on COCO streaming data. Checkpoint loading, COCO validation inference, proposal and detection visualization. ### YOLO from scratch (PyTorch) A five-notebook series building YOLOv8-style single-stage detection in pure PyTorch, from data loading through inference and evaluation. Single-stage detection design philosophy, anchor-free heads, and the YOLO architecture family. Streaming COCO, grid target assignment, and mosaic augmentation. CSPDarknet backbone with C2f bottleneck blocks. PANet feature pyramid neck and decoupled detection head. Task-aligned assignment, distribution focal loss, and training loop. NMS post-processing, COCO mAP evaluation, and latency benchmarks. ## Object segmentation Instance and semantic segmentation extend object detection to produce pixel-level masks, enabling fine-grained scene understanding. Extending Faster RCNN with a mask head for instance segmentation. Running inference with the TensorFlow Mask RCNN implementation. Visualizing COCO data loading, augmentation, and anchor generation. Layer-by-layer inspection of model activations and outputs. Visualizing learned filter weights and statistics. Detectron2 Mask RCNN training and evaluation workflow. Running pretrained Mask RCNN inference with torchvision. Encoder-decoder architecture for semantic segmentation. *** [Edit this page on GitHub](https://github.com/aegean-ai/eaia/edit/main/src/book/perception/index.mdx) or [file an issue](https://github.com/aegean-ai/eaia/issues/new/choose).