I don't understand the \"double counting\" comment on the Data tab regarding the redacted (i.e. black) parts of some images. What would be double counted?
\n\nFor example, from TrainSmall2 file, image 41.jpg has no black in the Train image but has a lot of redacted area in the TrainDotted. The blackened part of the TrainDotted has eliminated Sea Lions, yet the main Train image has them. In other words, the Train image could have 30 Sea Lions, but the number of dots in the respective TrainDotted could be 20.
\n\nDo we need to apply the redacted parts of the TrainDotted to the Train images so that we can correctly train our algorithms?
", "messages": [ { "id": "173629", "postDate": "04/07/2017 18:06:42", "content": "I don't understand the \"double counting\" comment on the Data tab regarding the redacted (i.e. black) parts of some images. What would be double counted?
\n\nFor example, from TrainSmall2 file, image 41.jpg has no black in the Train image but has a lot of redacted area in the TrainDotted. The blackened part of the TrainDotted has eliminated Sea Lions, yet the main Train image has them. In other words, the Train image could have 30 Sea Lions, but the number of dots in the respective TrainDotted could be 20.
\n\nDo we need to apply the redacted parts of the TrainDotted to the Train images so that we can correctly train our algorithms?
", "rawMarkdown": "I don't understand the \"double counting\" comment on the Data tab regarding the redacted (i.e. black) parts of some images. What would be double counted? \n\nFor example, from TrainSmall2 file, image 41.jpg has no black in the Train image but has a lot of redacted area in the TrainDotted. The blackened part of the TrainDotted has eliminated Sea Lions, yet the main Train image has them. In other words, the Train image could have 30 Sea Lions, but the number of dots in the respective TrainDotted could be 20. \n\nDo we need to apply the redacted parts of the TrainDotted to the Train images so that we can correctly train our algorithms?", "votes": null }, { "id": "174122", "postDate": "04/10/2017 11:50:21", "content": "I was wondering about the same and ask for clarification from the hosts.
\n\nThe mentioned image 41.jpg is a good example. Why is the blackened area distinct for Train and TrainDotted images? Which one is \"correct\" and should be used? This is very important to know since the number of sea lions differs in both images due to the redacted area.
", "rawMarkdown": "I was wondering about the same and ask for clarification from the hosts.\n\nThe mentioned image 41.jpg is a good example. Why is the blackened area distinct for Train and TrainDotted images? Which one is \"correct\" and should be used? This is very important to know since the number of sea lions differs in both images due to the redacted area.", "votes": null }, { "id": "174225", "postDate": "04/10/2017 18:47:59", "content": "If you check the test images, those also have black regions, so I'd superimpose the black regions of TrainDotted to Train and work with that.
", "rawMarkdown": "If you check the test images, those also have black regions, so I'd superimpose the black regions of TrainDotted to Train and work with that.", "votes": null }, { "id": "174522", "postDate": "04/11/2017 20:03:32", "content": "Yeah, I agree. I'm mostly curious what they don't want \"double counted\", which is apparently the motivation for the redacted regions.
", "rawMarkdown": "Yeah, I agree. I'm mostly curious what they don't want \"double counted\", which is apparently the motivation for the redacted regions.", "votes": null }, { "id": "179748", "postDate": "05/02/2017 18:31:43", "content": "These images are captured as we fly over and we have set the camera to fire so that there is adequate overlap between all images. This way we don't miss any animals. Since there is overlap, need to be sure that each sea lion is counted once in one images (that the same sea lion is not counted twice in overlapping images. Therefore, we manually look through images and black out areas that are already counted in another image (that is overlapping).
\n\nHope this helps!
", "rawMarkdown": "These images are captured as we fly over and we have set the camera to fire so that there is adequate overlap between all images. This way we don't miss any animals. Since there is overlap, need to be sure that each sea lion is counted once in one images (that the same sea lion is not counted twice in overlapping images. Therefore, we manually look through images and black out areas that are already counted in another image (that is overlapping).\n\nHope this helps!", "votes": null }, { "id": "179754", "postDate": "05/02/2017 18:52:13", "content": "The pictures is freedom your umanity peace and socializer people in the globe on sex religion nationality and dreames. Contact giacopal@gmail.com
", "rawMarkdown": "The pictures is freedom your umanity peace and socializer people in the globe on sex religion nationality and dreames. Contact giacopal@gmail.com", "votes": null } ], "comments": [ { "id": 174122, "author_name": "depthfirstsearch", "author_url": "", "post_date": "04/10/2017 11:50:21", "content": "I was wondering about the same and ask for clarification from the hosts.
\n\nThe mentioned image 41.jpg is a good example. Why is the blackened area distinct for Train and TrainDotted images? Which one is \"correct\" and should be used? This is very important to know since the number of sea lions differs in both images due to the redacted area.
", "votes": null, "replies": [ { "id": 174225, "author_name": "andraszsom", "author_url": "", "post_date": "04/10/2017 18:47:59", "content": "If you check the test images, those also have black regions, so I'd superimpose the black regions of TrainDotted to Train and work with that.
", "votes": null, "replies": [] }, { "id": 174522, "author_name": "mccroden", "author_url": "", "post_date": "04/11/2017 20:03:32", "content": "Yeah, I agree. I'm mostly curious what they don't want \"double counted\", which is apparently the motivation for the redacted regions.
", "votes": null, "replies": [] } ] }, { "id": 179748, "author_name": "sweenkl", "author_url": "", "post_date": "05/02/2017 18:31:43", "content": "These images are captured as we fly over and we have set the camera to fire so that there is adequate overlap between all images. This way we don't miss any animals. Since there is overlap, need to be sure that each sea lion is counted once in one images (that the same sea lion is not counted twice in overlapping images. Therefore, we manually look through images and black out areas that are already counted in another image (that is overlapping).
\n\nHope this helps!
", "votes": null, "replies": [] }, { "id": 179754, "author_name": "jimmy85", "author_url": "", "post_date": "05/02/2017 18:52:13", "content": "The pictures is freedom your umanity peace and socializer people in the globe on sex religion nationality and dreames. Contact giacopal@gmail.com
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "173629": "I don't understand the \"double counting\" comment on the Data tab regarding the redacted (i.e. black) parts of some images. What would be double counted? \n\nFor example, from TrainSmall2 file, image 41.jpg has no black in the Train image but has a lot of redacted area in the TrainDotted. The blackened part of the TrainDotted has eliminated Sea Lions, yet the main Train image has them. In other words, the Train image could have 30 Sea Lions, but the number of dots in the respective TrainDotted could be 20. \n\nDo we need to apply the redacted parts of the TrainDotted to the Train images so that we can correctly train our algorithms?", "174122": "I was wondering about the same and ask for clarification from the hosts.\n\nThe mentioned image 41.jpg is a good example. Why is the blackened area distinct for Train and TrainDotted images? Which one is \"correct\" and should be used? This is very important to know since the number of sea lions differs in both images due to the redacted area.", "174225": "If you check the test images, those also have black regions, so I'd superimpose the black regions of TrainDotted to Train and work with that.", "174522": "Yeah, I agree. I'm mostly curious what they don't want \"double counted\", which is apparently the motivation for the redacted regions.", "179748": "These images are captured as we fly over and we have set the camera to fire so that there is adequate overlap between all images. This way we don't miss any animals. Since there is overlap, need to be sure that each sea lion is counted once in one images (that the same sea lion is not counted twice in overlapping images. Therefore, we manually look through images and black out areas that are already counted in another image (that is overlapping).\n\nHope this helps!", "179754": "The pictures is freedom your umanity peace and socializer people in the globe on sex religion nationality and dreames. Contact giacopal@gmail.com" }, "source": "meta" }