Field Notes

source

NtechLab rebuttal of Cap able claims

StatusStableUpdatedStands on1 sourceSupportsnotes

In February 2023, Biometric Update reported that FR vendor NtechLab ran Cap_able’s own demonstration videos through its production recognition algorithms and successfully detected all faces, after Cap_able’s launch coverage claimed its adversarial knitwear renders wearers invisible to FR systems.

Cap_able had validated its garments against the online YOLO object detector. NtechLab’s response illustrates the two failure modes that recur in Adversarial clothing marketing:

  • an object detector is not a recognition pipeline, so passing YOLO does not support an FR-evasion claim;
  • a pattern fitted to one algorithm transfers poorly to another vendor’s system.

NtechLab is an interested party selling recognition systems, so its demonstration is not neutral either. It does not prove Cap_able garments never work; it proves the launch claim was validated against the wrong class of system. The episode is the clearest public case of an independent deployed-system check contradicting a vendor’s evasion marketing.

Built on 1 source (1 external).

Working out connections…