Abstract:
Identification of a snake to its species is very important, as it determines the
treatment to be provided to a snakebite victim. Administering the incorrect treatment
leads to detrimental effects and loss of life. Research has been conducted on using
computer vision for snake species identification; though it has not been done using
important external information to give a more accurate result. Previous research has
only used images during identification, and this leads to incorrect results, especially
when there are similar looking snake species in the country. In these cases, the location
and time of a sighting weight heavily in producing an accurate identification.
The solution proposed in this research identifies snakes by not just its image,
but external information in the form of the location and time of the snake sighting.
Using this information helped sway the identification of similar looking species to the
correct side. Near perfect identifications of the test set were achieved during
quantitative evaluation. Storing these identification results to display a real-time
distribution of these sightings, could be used by governmental authorities for a
multitude of reasons. This solution can be used in both critical scenarios, to geotag
live sightings, and educational purposes, as a simple image-only classification for
those interested in knowing more on the snake.
