Removing Human Bottlenecks in Bird Classification Using Camera Trap Images and Deep Learning

Round 1

Reviewer 1 Report

This is an interesting and well written paper. I can see the wider application to animal behaviour methods and data collection procedures across a wide range of situations. I recommend for publishing but have several minor comments for edit. 

I recommend for accessibility and ease of reading, to give the common names of species when their scientific name is introduced.

Remove the hyphen from eco-system services in the abstract. 

Line 104, is Aves really needed? Perhaps say avian taxonomy, given you have said mammal in the previous line. And perhaps change mammal to mammalian.

Line 135: does Wildlife Insights need a reference attached directly to it? Or a copyright or a website?

Line 166: why a challenge? Please evaluate. Again, I would just use birds or avian. 

Line 196 please provide a common name.

Figure 2: please explain what this figure shows in its caption.

Figure 3: the X axis is messy, due to the use of scientific names. Perhaps abbreviate to genus first letter and species, or use common name, or try to put a space between the genus and species names. Please provide lines for the Y and X axes. 

Please remove the capital letter from the species names in the X axis. 

Figure 4: why faster? Please explain or evaluate.

Figure 5: was there a specific battery used that could allow for continuous charging?

Figure 7: what is the relevance of this image? It seems to be advertising another platform of species identification. Can you provide evaluation of why it is included? Is the point to show that there is always a margin of error, because surely there is definitely more than a 73% chance of this animal being an elephant?!

Figure 9: Please consider format and bird names

Line 478: these are not classes, they are individual species. Please clarify.

Likewise, line 481 again mentions classes but these are species. Please clarify. 

Are you not referring to a taxonomic class?

Line 585; suggest "We present..."

Are you able to share the name of the technology and where it is available from, in the conclusion and at the end of the abstract? Has this research generated a tool that will be available for others to follow?

I appreciate birds were not manipulated in any way, but was this project ethically reviewed? You are collecting data on wildlife and you might capture other animals (including humans) on your footage. You might also disturb birds with the placement of equipment. Please provide some details, in your methods, on ethical review of aims, methods and experimental design. 

Excellent standard of written English. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

- The paper is well-written and well-structured. The organization of the paper follows a convenient natural flow that aids the reader in comprehendin the contribution. 

- Did the augmentation increase the data size?

- A concern of the work is that you did not do testing but rather you performed validation, which is not the same, and the results will not be reflective of true performance. 

- Figure 9. Absolute numbers are irrelevant to performance. 

- What feature extraction network (i.e., backbone did you use?)

- Provide the precision-recall curves. 

- Similar works utilizing Faster RCNN should be discussed and cited, see Khasawneh, N., Fraiwan, M. & Fraiwan, L. Detection of K-complexes in EEG waveform images using faster R-CNN and deep transfer learning. BMC Med Inform Decis Mak 22, 297 (2022). https://doi.org/10.1186/s12911-022-02042-x

 

This would confirm the validity and popularity of the methods and the choice of parameters. 

 

- The table of abbreviations is missing but maybe required by the journal.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf