- VGG → ResNet → DenseNet → EfficientNet
-
Why deeper ≠better without skip connections
-
Parameter-efficient models
-
Trade-offs: accuracy vs. speed vs. memory
Module 1 Computer Vision Foundations
The core concepts every real CV engineer must know before touching deep learning.
0/7
Module 2 Deep Learning for Vision
This module takes learners from basic image operations to full-scale deep learning pipelines for vision tasks. The focus is on understanding how CNNs extract hierarchical features, how models learn spatial patterns, and how to train, optimize, and deploy real-world vision models. Students build modern architectures from scratch, implement training loops, and gain the skills required to handle classification, detection, and segmentation workflows.
0/7
Module 3 Image Classification & Transfer Learning
Transfer Learning: Why it works
Fine-tuning vs Feature Extraction
Using Pretrained Models (ResNet, EfficientNet)
Improving real-world accuracy
Project: Build a high-accuracy classifier using transfer learning
Module 4 Object Detection
Traditional Detectors (Haar, HOG + SVM)
Deep Learning Detectors
YOLO family
SSD
Faster R-CNN
Bounding box loss functions
NMS & anchor boxes
Project: Train a YOLO object detector on a custom dataset
Module 5 Image Segmentation
Sematic vs Instance vs Panoptic segmentation
U-Net Architecture Breakdown
Mask R-CNN
Practical tips for dataset annotation
Project: Build a semantic segmentation model for medical or industrial images