I generated a "stress test" of 200 rare defects from 7 real photos
Details
- External ID
- 46994542
- Source
- HN
- Company
- —
- Product
- —
- Website domain
- —
- Launched
- Feb. 12, 2026
- Cohort
- —
- Upvotes
- 6
- Upvotes percentile
- 0.28099730458221023
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:25 p.m.
- Updated at
- Sept. 7, 2026, 9:25 p.m.
Description
Hello HN,I work on vision systems for structural inspection. A common pain point is usually that while we have a lot of "healthy" images, we often lack a reliable "Golden Set" of rare failures (like shattered porcelain) to validate our models before deployment.You can't trust your model's recall if your test set only has 5 examples of the failure mode for example.So to fix this, I built a pipeline to generate datasets. In this example, I took 7 real-world defect samples, extracted their topology/texture, and procedurally generated 200 hard-to-detect variations across different lighting and backgrounds.I’m releasing this batch of broken insulators (CC0) specifically to help teams benchmark their model's recall on rare classes:https://www.silera.ai/blog/free-200-broken-insulators-datase...- Input: 7 real samples.- Output: 200 fully labeled evaluation images (COCO/YOLO).- Use Case: Validation / Test Set (not full training).How do you guys currently validate recall for "1 in 10,000" edge cases?Jérôme
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Manufacturing
- Function
- Content generation
- Audience
- B2B
- AI stance
- AI feature
- Project type
- Commercial product
- Normalized one-liner
- defect generation from photos
- Manually corrected
- False
Could you build this?
Partial Building a script to prompt diffusion models is trivial, but producing photorealistic, physically accurate, domain-specific structural defects for training and validating industrial computer vision systems requires specialized generative CV pipelines and synthetic data control.
What it would actually take: The architecture relies on fine-tuned diffusion models (e.g., Stable Diffusion XL or Flux) using LoRA or ControlNet conditioned on depth maps, edge detection, and material properties. The hard part is generating photorealistic, physically plausible defect textures (fractures, spalling, microcracks) that match the precise optical characteristics of real industrial inspection cameras without domain drift or synthetic artifacts, requiring expertise in CV data synthesis and quality auditing.
Discussion
4 comments analyzed.
Concerns raised: Requires registration and credit purchase to download free dataset, Data access feels like email trap/spam collection, Broken 'Get Dataset' button after signup page navigation
Feature requests: Direct download link instead of registration requirement, Track downloads via server logs
Competitors
Other products that read as similar to this one — 16 launches clear the similarity bar, closest 8 shown.
Attention rank: #14 of 17 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 41 days after the earliest competitor.
- Jacobi–IDE for Abaqus subroutine with analytical tests and AI diagnosis · hn · 2026-06-21 · 18 upvotes · similarity 0.37
- Bulk Image Generator · hn · 2026-03-07 · 5 upvotes · similarity 0.34
- SmartCheck · ph · 2026-09-18 · 57 upvotes · similarity 0.34
- The Two by Two Truth Diagram · hn · 2026-03-21 · 5 upvotes · similarity 0.33
- I wrote down every expensive hardware development mistake I've seen · hn · 2026-03-08 · 11 upvotes · similarity 0.33
- Synthetic Webcam Scene Dataset · ph · 2026-09-30 · 4 upvotes · similarity 0.33
- Memory Test · ph · 2026-09-23 · 1 upvotes · similarity 0.32
- I built a reference site for the recurring hard problems in software · hn · 2026-04-27 · 9 upvotes · similarity 0.32
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a content generation tool for Government yet.