Computer Vision Interview 20 essential Q&A Updated 2026
Face

Face Recognition: 20 Essential Q&A

Detect, align, embed, and match—biometric pipelines and modern metric learning.

~11 min read 20 questions Advanced
MTCNNembeddingtripletArcFace
Quick Navigation
1. Face recognition pipeline?
2. Face detection
3. Alignment
4. Embeddings
5. Verification
6. Identification
7. Triplet loss
8. ArcFace / CosFace
9. FaceNet idea
10. LFW / benchmarks
11. Threshold selection
12. Anti-spoofing
13. Masks / occlusion
14. Demographic bias
15. Privacy
16. 3D face / pose
17. On-device
18. Image quality
19. Synthetic data
20. Presentation attacks
1 Typical pipeline? ⚡ easy
Answer: Detect face → align to canonical pose → CNN embedding → compare cosine/L2 distance.
2 Face detection? 📊 medium
Answer: Find boxes/scales (MTCNN, RetinaFace, YuNet)—must handle profile, small faces, and clutter before recognition.
3 Alignment? 📊 medium
Answer: Use 5 or more landmarks to similarity-transform face to fixed template—reduces pose/light variance before embedding.
4 What is an embedding? 📊 medium
Answer: L2-normalized vector (e.g. 512-D) such that same identity is close, different identities far—learned with metric objectives. 
sim = F.cosine_similarity(emb_a, emb_b)  # face verification
5 Verification vs identification? ⚡ easy
Answer: Verification: same person or not (1:1). Identification: match probe to gallery (1:N)—needs threshold and rank metrics.
6 Open-set identification? 📊 medium
Answer: Probe may be unknown—need rejection option based on similarity threshold to avoid false accepts.
7 Triplet loss? 🔥 hard
Answer: Anchor closer to positive than to negative by margin—hard negative mining critical for convergence (FaceNet).
8 ArcFace? 🔥 hard
Answer: Angular margin on hypersphere between logits—enforces larger inter-class angular separation; state-of-the-art metric learning.
9 FaceNet? 📊 medium
Answer: End-to-end CNN with triplet loss producing compact embeddings—popularized deep face recognition at scale.
10 Benchmarks? ⚡ easy
Answer: LFW, CFP-FP, IJB-C, MegaFace—vary in pose, N protocol, and difficulty; report TAR@FAR for verification.
11 Threshold tuning? 📊 medium
Answer: Set operating point on validation to balance FAR vs FRR for the deployment constraint (access control vs convenience).
12 Anti-spoofing? 📊 medium
Answer: Detect print/screen/replay attacks with texture, depth, or rPPG—required for liveness in banking kiosks.
13 Masks / COVID era? 📊 medium
Answer: Periocular focus, synthetic mask augmentation, or dedicated training—lower accuracy if model not adapted.
14 Demographic bias? 🔥 hard
Answer: Unequal error rates across groups—audit datasets, balanced training, and fairness constraints in deployment.
15 Privacy? ⚡ easy
Answer: Biometric data is sensitive—encrypt templates, consent, retention limits, on-device processing where possible.
16 3D morphable models? 📊 medium
Answer: Fit 3DMM for pose-invariant recognition or generate synthetic views—helps extreme pose.
17 On-device? ⚡ easy
Answer: Quantized MobileFaceNet-style backbones, NNAPI/CoreML—latency and power constrained.
18 Quality assessment? 📊 medium
Answer: Blur, exposure, resolution gates before embedding—reject low-quality captures to reduce false matches.
19 Synthetic faces? 📊 medium
Answer: GAN-generated diversity for training—watch for domain gap and identity leakage in synthetic sets.
20 Presentation attacks? 📊 medium
Answer: ISO standards categorize attack instruments—multimodal liveness (depth, IR) mitigates many.

Face Recognition Cheat Sheet

Steps
  • Detect → align
Metric
  • Cosine / L2
  • ArcFace
Ops
  • FAR/FRR

💡 Pro tip: Align landmarks; normalize embeddings; tune FAR vs FRR.

Full tutorial track

Go deeper with the matching tutorial chapter and code examples.

Face Recognition Tutorial
Previous Next