Computer Vision
CSE471Prof. Makarand Tapaswi + Prof. Charu Sharma•Spring 2025-26•4 credits
Mock Paper 14 — Comparison-Only ('X vs Y' / 'X vs Y vs Z')
Duration: 150 min • Max marks: 100
Section A — Quick Comparisons (2 marks each, 40 marks)
40 marks- 1.Erosion vs Dilation — one-line distinction.2 m
- 2.Opening vs Closing.2 m
- 3.Mean vs Median filter.2 m
- 4.Sobel vs Laplacian.2 m
- 5.DCT vs DFT for image compression.2 m
- 6.RGB vs HSV colour space.2 m
- 7.RGB vs Lab colour space.2 m
- 8.Histogram equalisation vs Gamma correction.2 m
- 9.Faster R-CNN RPN vs Selective Search.2 m
- 10.One-stage (YOLO) vs Two-stage (Faster R-CNN) detection.2 m
- 11.Anchor-based vs anchor-free detection.2 m
- 12.RoI Pool vs RoI Align.2 m
- 13.Semantic vs Instance vs Panoptic segmentation.2 m
- 14.FCN vs U-Net for segmentation.2 m
- 15.ResNet skip vs U-Net skip.2 m
- 16.BatchNorm vs LayerNorm.2 m
- 17.Pre-norm vs Post-norm Transformer.2 m
- 18.Self-attention vs Cross-attention.2 m
- 19.Absolute vs Relative vs Rotary positional encoding.2 m
- 20.CLIP vs SigLIP loss.2 m
Section B — Detailed Comparisons (5 marks each, 30 marks)
30 marks- 1.SimCLR vs MoCo vs DINO — compare on (a) negatives, (b) projection head, (c) what stabilises training.5 m
- 2.Generative vs Discriminative representations — VAE vs MAE vs GAN vs Diffusion.5 m
- 3.PointNet vs PointNet++ vs DGCNN for point clouds.5 m
- 4.SGD vs SGD+Momentum vs Adam vs AdamW.5 m
- 5.NeRF vs 3D Gaussian Splatting.5 m
- 6.CLIP vs DINO vs MAE vs JEPA for downstream tasks.5 m
Section C — Synthesis Comparisons (10 marks each, 30 marks)
30 marks- 1.Compare R-CNN, Fast R-CNN, Faster R-CNN, YOLO, DETR on: proposal mechanism, feature sharing, anchors, end-to-end, speed, strengths/weaknesses.10 m
- 2.Compare VGG, ResNet, Inception, MobileNet, EfficientNet, ViT, ConvNeXt as image-classification backbones.10 m
- 3.Compare PaliGemma, BLIP-2, LLaVA, Qwen2-VL, GPT-4V as VLMs.10 m
Track your attempt locally — score and time are recorded in your browser. (Coming soon: timed-attempt mode.)