BSDA5006 — Deep Learning for Computer Vision (DL-CV)

247 words

1 min read

Visual companion

Python

Type and operator map

Python Week 1: the first filter for runtime behavior

Open gallery Related surface

View

Revision summary

What this note is really saying

Short form

# BSDA5006 — Deep Learning for Computer Vision (DL-CV) > **Course**: Degree Elective (4 credits) **Topics**: CNNs, Advanced architectures, Object detection, Segmentation, GANs, Vision Transformers **Files**: 10 comprehensive topic files ## Course Index File Topic Key Concepts 01-cnn-fundamentals CNN Architecture Con...

BSDA5006 — Deep Learning for Computer Vision (DL-CV)

Course: Degree Elective (4 credits) Topics: CNNs, Advanced architectures, Object detection, Segmentation, GANs, Vision Transformers Files: 10 comprehensive topic files

Course Index

File	Topic	Key Concepts
01-cnn-fundamentals	CNN Architecture	Convolution, pooling, stride, padding, feature maps, receptive field
02-advanced-cnns	Advanced CNNs	VGG, Inception, ResNet, DenseNet, MobileNet, EfficientNet
03-transfer-learning	Transfer Learning	Pre-training, fine-tuning, feature extraction, domain adaptation
04-object-detection	Object Detection	R-CNN, Fast R-CNN, YOLO, SSD, anchor boxes, IoU, NMS, mAP
05-segmentation	Segmentation	Semantic, instance, panoptic, FCN, U-Net, Mask R-CNN
06-gans	GANs	Generator, discriminator, DCGAN, CGAN, WGAN, CycleGAN
07-vision-transformers	Vision Transformers	ViT, patch embedding, positional encoding, hybrid architectures
08-data-augmentation	Data Augmentation	Transformations, MixUp, CutMix, RandAugment, AutoAugment
09-image-processing	Image Processing	Gradients, filtering, edge detection, frequency domain, histograms
10-diffusion-models	Diffusion Models	DDPM, noise scheduling, U-Net for denoising, sampling

Key Formulas

Formula	Description
$O = (W - K + 2P)/S + 1$	Output size after convolution
$\text{Receptive Field} = 1 + \sum (K_i - 1)$	RF computation for stacked conv layers
$\text{IoU} = \frac{TP}{TP + FP + FN}$	Intersection over Union
$mAP = \frac{1}{N}\sum AP$	Mean Average Precision
$\mathcal{L}_{GAN} = \min_G\max_D [\log D(x) + \log(1-D(G(z)))]$	GAN loss
$\text{Patch Embedding}(I) = \text{Linear}(\text{Concat}(p_1, p_2, ..., p_n))$	ViT patch embedding

Exam Weightage

Topic	Quiz 1	Quiz 2	End Term
Image Processing	★★★★★	★★★	★★
CNN Fundamentals	★★★★★	★★★★	★★★
Advanced CNNs	★★	★★★★★	★★★★
Transfer Learning	★★	★★★★★	★★★
Object Detection	★	★★★	★★★★★
Segmentation	★	★★★	★★★★★
GANs	★	★★★	★★★★
Vision Transformers	★	★★	★★★★

Cross-Course Links

BSDA5004 (LLMs): Transformer attention is shared with ViTs
BSDA5002 (GenAI Foundations): GANs and diffusion models covered in-depth
BSDA5013 (DL Practice): Hands-on CV implementation projects Join Discord PreviousDiffusion Models for CV