Statistics

Recognition rate

AmphIdent reaches a very high recognition rate of about 98%. The following diagrams and tables show further information.

Crested newts

The following diagram visualizes the relation between size of the database and the recognition rate.

Pattern recognition performance of individual photo-identification for crested newts

With a database size of 2000 individuals you need to compare 7 animals manually to reach a recognition rate of 98%. 95% of all assignments can be found with the comparison of only one pattern with a database size of 2000 individuals.

Extrapolating the course of the lines reveals that 98% of all assignments can be found by manual comparison of 10 patterns at a database size of 3000 individuals. 97% of all assignments will be even found with much larger databases.

Further investigations on the recognition rate for crested newts can be found in:

AXEL DRECHSLER, TOBIAS HELLING & SEBASTIAN STEINFARTZ: Genetic fingerprinting proves cross-correlated automatic photo-identification of individuals as highly efficient in large capture-mark-recapture studies in "Ecology and Evolution (2015)", Volume 5, Issue 1, pp. 141-151 Link

Duration of processing one picture

The processing of one photo divides into two steps:

  1. Create a pattern from the picture
  2. Comparison of this pattern with already existing ones.

The following table explains the time needed for processing pictures of different species:

Species Duration for pattern creation Duration for comparison of the pattern Overall time*
Crested newt ~2min 1ms ~5min
Fire-bellied toad ~1min 1ms ~5min

* The time needed for comparison of one pattern with another. The value bases on a Pentium 4/3GHz.

**The shown Overall time approximates the duration of processing and comparison at a database size of 2500 patterns. For the final manual comparison two seconds per pattern were assumed.

The use of a faster PC will shorten the duration remarkably. With large databases the comparison is the most time-consuming task. Running a multi-core system will accelerate the process even further.