VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding
u_Unrolled
D18: Multimodal Foundation Models · Vision in Production — A Defect-Detection Project with YOLO, and Serving Models on the Edge

Vision in Production — A Defect-Detection Project with YOLO, and Serving Models on the Edge

10 min read·Advanced· Unrolled Pro
the project pipeline (capture → label with bounding boxes → train a one-stage detector → evaluate mAP and per-class recall at the operating threshold → export to ONNX → quantize to INT8 → deploy on an edge device with a frame-skipping loop), with a latency budget bar for one frame (pre-process, inference, NMS, post-process) against the 66 ms deadline for 15 fps.
Diagram— the project pipeline (capture → label with bounding boxes → train a one-stage detector → evaluate mAP and per-class recall at the operating threshold → export to ONNX → quantize to INT8 → deploy on an edge device with a frame-skipping loop), with a latency budget bar for one frame (pre-process, inference, NMS, post-process) against the 66 ms deadline for 15 fps.

You already know: detection, NMS, IoU and mAP (Computer vision); CNNs and their cost (CNNs from scratch); quantization formats and calibration (D12 Quantization).

Why this matters for Atlas. Most real computer vision is narrow, specific and runs close to the camera: inspecting parts, counting items, spotting failures. Success depends less on the newest architecture than on data (rare defects, lighting changes), on an evaluation that reflects what an error costs (a missed failure wastes a 10-hour print; a false alarm interrupts an operator), and on fitting a latency and memory budget on cheap hardware. This lesson walks one project end to end: data and labels, a one-stage detector (the YOLO family), evaluation and thresholding, and edge deployment with export, quantization and a real-time loop.


1. Data and labels: the real work

The question. Where do the training images for rare defects come from?

Start from nothing. Defects are rare, so collect deliberately: run prints until they fail, include the conditions the camera will really see (lighting, angles, filament colours, dust on the lens), and label with bounding boxes per defect class ("spaghetti", "layer shift", "detached corner"). Measure label quality: have two people label a sample and compute their box agreement (IoU) and class agreement; inconsistent labels cap the model's achievable mAP. Split by print job, not by frame, because frames from the same print are near-duplicates (the same leakage as splitting churn data by row). Augment with what changes in deployment (brightness, blur, small rotations), not with transformations that can't happen.

That's the preview. The full lesson, with its code and exercises, is part of Unrolled Pro.
Staff AI & Pro Tier Lesson

Unlock "Vision in Production — A Defect-Detection Project with YOLO, and Serving Models on the Edge"

The foundations are free forever. Unrolled Pro unlocks the complete lesson, production architectures, and 55 battle-tested ML system design case studies.

Everything included with Pro:
All 234+ Deep-Dive Lessons across transformer internals, inference, RAG & agents
55 Staff-Level ML & System Design Case Studies with mathematical capacity estimates
AI Engineer Sheet & Exercises with worked solutions, self-checks, and code
100 Landmark Research Papers guided walkthroughs & progress tracking
14-Day Money-Back Guarantee — 100% full refund if you aren't satisfied
View Plans & Unlock All Lessons

From ₹167/mo (~$2/mo) · Cancel anytime · Instant activation

100% Risk-Free: 14-day no-questions-asked money-back guarantee