Preparing Samples for Machine Vision Evaluation

What Samples to Send

You need three categories of samples, each clearly labeled and separated:

Good parts (100-200 pieces): Representative of your normal production output. These define the acceptance baseline. If your good parts have acceptable variation, such as normal plating color range or minor surface texture, include examples spanning that range. Do not send only perfect parts, send real-world acceptable parts.

Defective parts (30-50 pieces per defect type): Every defect type you need the machine to catch. Label each defect type separately: missing thread, head deformation, surface scratch, plating void, crack, dimensional out-of-spec, mixed material. If you do not have enough physical defective samples, include close-up photos with annotations.

Borderline parts (20-30 pieces): Parts that are on the edge of accept/reject. These are the hardest to classify and the most important for tuning the system. If the machine can correctly handle borderline parts, it will handle normal production with confidence.

How to Pack and Ship

Pack each sample category in separate bags or containers, clearly labeled. Use rigid containers for delicate parts to prevent damage during shipping. Include a packing list with part name, part number, quantity per category, and defect type labels.

Ship to our Dongguan test lab. We typically complete evaluation within 5-7 working days of sample receipt. The test report includes: detection rate per defect type, false reject rate on good parts, borderline part classification results, recommended machine configuration, throughput estimate, and sample images showing the machine’s view of each defect type.