Review Reports

Round 1

Reviewer 1 Report

Dear Editor and Authors, the work is surely of interest of Remote Sensing readers. It is well written, even though terminology must be improved. For instance, litter is not composed only by plastics, and overall the paper must specify that the work regards "floating" litter.

M ore details can be added in Methods and Results. More importantly, visualization must be much improved: most of figures are not explanatory of the work. I suggest specific improvements below, but in general it is necessary to show the monitored area and increase the number of sample images with annotations of the items. Discussion can also compare this work with more works: there are plenty of papers that already develoepd different algortihms for litter detection (both floating and beached) with different techniques and overall good assessments. 

Overall, a nice work that deserves to be published after some improvements!

To Authors.

35-36 please, specify the topic is "floating" litter

39 marine litter is not composed only by plastic. Overall, plastic is always about 80%, but other materials can be found.

64-65 also multispectral https://doi.org/10.1016/j.marpolbul.2022.113431

65-67 not totally true. The works of Garcia-Garin et al., for instance, already compared ground-truth against drone-based surveys. I suggest to re-write the paragraph, distinguishing among sinthetic experiments (e.g., satellite) and operational surveys. https://doi.org/10.1016/j.envpol.2019.113680.

80-81 I wouldn't say "significantly". The surveys were quite successfull, yet, sun glare can also affect on board cameras.

Fig.1-2 please, add also a map of the study site and the size of the monitored areas. Images can be joined togehter in a panel, it is not necessary to show boat and cameras with such big pictures. Please, help the reader to identify the items you indicate in captions: arrows and/or text and/or boxes can be drawn on images.

In the text, please specify the time needed for the survey

Fig.3 please, specify in caption the meaning of "positive" and "negative" images.

Fig.5 and section 2.4 not clear to me the pixel percentage approach. First, image can be easily pre-processed to crop the useless portion (for instance, sky). This would fast the processing and lower the percentage of detected pixel, so the informatin is not so explanatory. Please, try to improve the description of pixel-based approach: in terms of litter survey, it is more important the number of objects (and their size) than the number of pixels.

In Methods, I suggest to improve the image quality and number of image examples.

240-248 any other type of objects found? were they considered false positive?

256-259 plastic bottles? transparent? it is relevant to specify it, since some items can be difficult to detect due to their transparency

Fig.7 please, improve the number of examples, and mark on images the items. Nothing can be seen in the provided image

3.2.1 same comment as above. In the end, it is not clear the number of items detected and or missed. Try please to improve the table in this regard

272 still, please specify "floating"

283-286 also, the work of Kylili (correct the name please) https://doi.org/10.1007/s11356-019-05148-4 was an exercise of machine learning, and not an actual survey as your work. 

316 319 I do not totally agree. The object-based approach also allows to count the number of items and estimate their size, which is fundamental for a proper litter monitoring task.

322-324 awkward to read the sentence. Surely, assessments are reasonale, but as said before, the actual area of interest can be simply reduced. Also, simple machine learning can be based on colours, see perfomances of other algorithms in https://doi.org/10.1016/j.marpolbul.2021.112594

325-327 exactly. And also computational time

346 "quantify"? there is not an actual survey of floating litter and counts of items. Also, it is not mentioned if the automated algortihm might have counted the same items over the image series, please specify that.

357-361 you can also propose the use of multispectral camera

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

In general, this study is very interesting to analyse the impact of marine debris. But it is important to explain better the method and the model wwith results obtained. Please, insert a map or cartography of study area. The model is explained but it is important to link the test of the model with results (number of plastic debris and other). 

Line 120: Please, add a parapraph with description of study area with cartographies and some images. Moreover, add a brief description of natural features of study area, wind, marine corrents and so on, that can influence the presence and mouvement of debris in the sea. 

Line 268: in the Results paragraph add a table with a total of debris found during the anlysis; that is a clarification of plastic materials found using this model in the study area.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

I guess it is an interesting and novelty work well presented and referenced.

Anyway I am not an expert on monitoring systems so I have just a few questions/suggestions.

Some references are not presented in the right way, see line 101 (Tata et al.2021)” and 104 (Van Lieshot et al.2020), you need to add number after them….the same at lines 286 and 288 and 306…

Line 125. “shopping bags (J3)” and following lines….what are the codes for?

 

Line 134, “October-November 2021”…Could you please indicate the period investigated? was it all the October/November period? Or few days within that time span?

 

Lines 140-141, please clarify how many hours of observation were carried out each day.

 

Line 145, what about water turbidity? Is it a relevant problem or not?

 

Lines 242-246, very interesting, could you please give more details?

 

Figure 7. I am sorry I am not able to see the bottles…..could be possible to make a zoom on them? I mean you can present the figure as it is and than add a zoom.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

I have reviewed the manuscript entitled "Detection and Classification of Floating Plastic Litter Using a Vessel-Mounted Video Camera and Deep Learning" by Sophie Armitage et al. for publication in Remote sensing. The authors present an automated approach to collecting in situ data using a trained object-detection algorithm to detect and quantify marine macroplastics from video footage taken from vessel-mounted camera. The manuscript is well organized, and the subject matter is very interesting, so I recommend it for publication after minor revisions.

I suggest highlighting more the novelty of the work done than what is already in the literature and improve the introduction and the discussion of results (i.e. consider the following references: Sannigrahi et al. 2012, Development of automated marine floating plastic detection system using Sentinel-2 imagery and machine learning models, https://doi.org/10.1016/j.marpolbul.2022.113527; Gonçalves et al. Operational use of multispectral images for macro-litter mapping and categorization by Unmanned Aerial Vehicle, https://doi.org/10.1016/j.marpolbul.2022.113431). Add some aspects regarding the possibility to detect microplastics and individuate the type of polymers with the employed system compared to classic detection (see Microplastics in the Aquatic Environment: Occurrence, Persistence, Analysis, and Human Exposure, https://doi.org/10.3390/w13070973).

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Dear Editor and Authors, the revised manuscript improved the previous version. My requirements have been satisfied, so in my opinion the paper can be published. 

Still missing one more thing, that needs to be clarified: in Methods and Table 1, wind speed is reported. While it is soecified that pictures were taken when boat was moored in some sites, it is not indicated what was the boat speed offshore. This may be of interest for replicating the work: boat speed can influence the quality of image and video frames.  

Congrats on the work.