Appendix C β Glossary
βA shared vocabulary is the foundation of shared understanding.β
Notation Reference
Planned content: Tensor dimension notation used throughout the guide. Convention:
[B, C, H, W]for image batches (B=batch, C=channels, H=height, W=width).[B, T, D]for sequences (T=time/length, D=dimension).[N, 4]for bounding boxes (N=number of boxes, 4=coordinates). Superscripts for layer index. Subscripts for channel/class index.
π Planned table: Notation lookup β symbol, meaning, typical shape, where used.
AβC
Planned content: AnchorBox, ASPP (Atrous Spatial Pyramid Pooling), ArcFace, AutoAugment, Average Precision (AP), Backbone, Batch Normalization (BN), Bilinear Interpolation, Bounding Box (bbox), BYOL, Camera Intrinsics, Class Activation Map (CAM), Classification Head, CLIP, CLS Token, Confusion Matrix, Contrastive Loss, ConvNeXt, CosFace, Cross-Entropy Loss, CTC (Connectionist Temporal Classification), CutMix.
DβF
Planned content: Data Augmentation, Deep Supervision, DETR, Dice Loss, Dilated Convolution, DINOv2, Disparity, Distillation, DPM (Deformable Part Model), DW-Separable Convolution, EfficientNet, Embedding, EMA (Exponential Moving Average), Encoder-Decoder, EPE (Endpoint Error), FID (FrΓ©chet Inception Distance), FPN (Feature Pyramid Network), Focal Loss, FLOP.
GβK
Planned content: GAP (Global Average Pooling), GAN (Generative Adversarial Network), GIoU, GradCAM, Grouped Convolution, Hard Mining, Heatmap (keypoint), HOG (Histogram of Oriented Gradients), HOTA (Higher Order Tracking Accuracy), HRNet, Hungarian Algorithm, IDF1, Instance Normalization, Instance Segmentation, IoU (Intersection over Union), Kalman Filter.
LβN
Planned content: Label Smoothing, LiDAR, Linear Probing, LoRA, Long-Tail Distribution, LovΓ‘sz Loss, MAE (Masked Autoencoder), mAP (mean Average Precision), MixUp, MobileFaceNet, MobileNet, Momentum Encoder, MOTA (Multiple Object Tracking Accuracy), MPJPE, Multi-Scale Training, NeRF, NMS (Non-Maximum Suppression), NTXent (Normalized Temperature-scaled Cross Entropy).
OβR
Planned content: OKS (Object Keypoint Similarity), One-Stage Detector, ONNX, Open-Set Recognition, OpenPose, PAF (Part Affinity Field), Panoptic Segmentation, Patch Embedding, PointNet, PointPillars, Positional Encoding, Pseudo-Label, PTQ (Post-Training Quantization), QAT (Quantization-Aware Training), ReLU, Re-ID (Re-Identification), ResNet, RoIAlign, RoIPool, RPN (Region Proposal Network).
SβZ
Planned content: SAM (Segment Anything Model), Semantic Segmentation, SimCLR, SIFT, Skip Connection, Soft-NMS, Softmax, Sparse Convolution, SphereFace, STE (Straight-Through Estimator), Stride, Swin Transformer, TAR@FAR, TensorRT, Timm, Top-Down Pose, Triplet Loss, Two-Stage Detector, U-Net, ViT (Vision Transformer), VQ-VAE, WBF (Weighted Box Fusion), YOLO.
Acronym Quick Reference
Planned content: Single-page table β acronym, full name, chapter reference.
π Planned table: Full acronym table β abbreviation β full name β chapter(s) where discussed.
Last updated: May 2026