How it Works
The user has to upload the field image of the butterfly or moth in question along with the date and location of capture (optional). The image uploaded by the user will be compared with the simulated models of the Artificial Intelligence Neural Network System to best fit it to the existing trained datasets. The best-fit match to the sample butterfly will be shown to the user with a confidence level in percentage. In the end, the user can choose most appropriate suggestion.
Note: The confidence percentage of each suggestion is estimated for 100% (all the three suggestion do not sum up to 100%).
- The programme is trained to identify the butterflies and moths as a species (not larva, pupa, eggs, male or female and sub-species) in its natural environment and hence the background does not influence on the identification (refer image a).
- In case the butterfly occupies a small portion of the image canvas it is suggested to crop the image (as close as shown in the first image) for the best result (refer image b)
- Cropping too close to the butterfly may end up in poor identification percentage (refer image c)
- In case of multiple butterflies the programme identifies the butterfly randomly from the image. However, it is advised to crop and submit the image for better result (refer image d)
- In case of poor identification percentage user can redo cropping or uncropped images for better identification.
- The software do not recognise any image other than butterfly or moth in it.
- If sample is identified with less confidence percent it is suggests that the submitted image might be blur, or captured in uncommon angles or the species may be new to the trained dataset (other than 1011 butterflies and 1111 moths listed)
a. Butterfly in natural condition
b. Wide area image
c. Close cropping
d. Multiple butterflies in a single image
For any queries or report error please visit the contact section.