Practice — Case Study: Visual Defect Detection on a Production Line (6 questions)

A pharma blister-pack line has these cost parameters: cost of an escape (a defective blister reaching a pharmacy) C_escape = $2,000 (recall exposure, regulatory reporting cost, amortized brand risk); cost of a false reject C_scrap = $3 (material plus rework). The line produces 40,000 blisters/day and the plant has a hard scrap-budget ceiling of 1.5% of daily output.

1. Compute the break-even probability p* using reject if p(s)·C_escape > (1-p(s))·C_scrap.
2. Your anomaly detector's score distribution on golden (normal) parts has a long right tail: roughly 2.5% of genuinely normal parts score above the level that corresponds to p(s) = p*. What does this mean for the reject policy, and how do you resolve the conflict with the scrap-budget ceiling?
3. Propose a concrete three-band policy (pass / review / reject) that respects both the cost-derived threshold and the scrap budget, and explain what happens to the parts that would have been auto-rejected under the pure cost rule but can't be under the scrap budget.

You're onboarding two new SKUs onto edge boxes that already host several other SKUs' models:

• SKU A is a textured fabric panel where defects are subtle, diffuse weave irregularities spread across large regions, not sharply-localized marks. The edge box for this line is a higher-spec unit with generous RAM headroom.
• SKU B is a small metal bracket with sharp, highly localized defects (a crack, a missing hole). The edge box for this line is already close to its memory ceiling from the other SKUs it hosts, and adding one more large PatchCore-style memory bank would exceed it.

1. Which approach-ladder rung (reconstruction-based / GAN, or feature-memory-bank) would you argue for on each SKU, and why?
2. For SKU B specifically, name the concrete architectural reason a memory-bank method's footprint grows in a way a trained generator's doesn't, and what lever (from the case) you'd pull before abandoning the memory-bank approach entirely.
3. If you picked the reconstruction rung for either SKU, which of AnoGAN or f-AnoGAN would you deploy, and why does the other one fail this line's constraints outright?