how is everyone using all the masks available, there are some 4k+ mask, do we add all of them into the picture? or how are we doing it, wouldnt that require a lot of memory as every image would have 4k+ images to deal with
", "messages": [ { "id": "1754159", "postDate": "04/13/2022 12:23:20", "content": "how is everyone using all the masks available, there are some 4k+ mask, do we add all of them into the picture? or how are we doing it, wouldnt that require a lot of memory as every image would have 4k+ images to deal with
", "rawMarkdown": "how is everyone using all the masks available, there are some 4k+ mask, do we add all of them into the picture? or how are we doing it, wouldnt that require a lot of memory as every image would have 4k+ images to deal with", "votes": null }, { "id": "1754308", "postDate": "04/13/2022 14:31:25", "content": "You should check out How do you use the masks provided in the dataset? post.
\nThe masks in train_masks folder are masks for test_images (the folder name is not correct so it's a little confusing).
Abby Stylianou [Competition Host]:
\n\n\nThe mask files are available for your convenience -- if there's a query image called 0001.jpg and a mask called 0001.png, the mask is simply a PNG that includes the exact mask from the query JPG. You can use these for processing of the query images in case it's easier than detecting the mask from the JPG, and also may also use any of the masks for whatever purpose in your training
\n
So when doing predictions on the test dataset if there is image 0001.jpg and existing file with name 0001.png in the train_masks folder you can use the mask image to select the occlusion in the test image. It's just to make it easier to handle but you don't have to use them. I am currently not using them in any way and for training just generate random occlusion.
", "rawMarkdown": "You should check out [How do you use the masks provided in the dataset?](https://www.kaggle.com/competitions/hotel-id-to-combat-human-trafficking-2022-fgvc9/discussion/313547) post. \nThe masks in train_masks folder are masks for test_images (the folder name is not correct so it's a little confusing).\n\nAbby Stylianou [Competition Host]:\n> The mask files are available for your convenience -- if there's a query image called 0001.jpg and a mask called 0001.png, the mask is simply a PNG that includes the exact mask from the query JPG. You can use these for processing of the query images in case it's easier than detecting the mask from the JPG, and also may also use any of the masks for whatever purpose in your training\n\nSo when doing predictions on the test dataset if there is image 0001.jpg and existing file with name 0001.png in the train_masks folder you can use the mask image to select the occlusion in the test image. It's just to make it easier to handle but you don't have to use them. I am currently not using them in any way and for training just generate random occlusion.", "votes": null } ], "comments": [ { "id": 1754308, "author_name": "michaln", "author_url": "", "post_date": "04/13/2022 14:31:25", "content": "You should check out How do you use the masks provided in the dataset? post.
\nThe masks in train_masks folder are masks for test_images (the folder name is not correct so it's a little confusing).
Abby Stylianou [Competition Host]:
\n\n\nThe mask files are available for your convenience -- if there's a query image called 0001.jpg and a mask called 0001.png, the mask is simply a PNG that includes the exact mask from the query JPG. You can use these for processing of the query images in case it's easier than detecting the mask from the JPG, and also may also use any of the masks for whatever purpose in your training
\n
So when doing predictions on the test dataset if there is image 0001.jpg and existing file with name 0001.png in the train_masks folder you can use the mask image to select the occlusion in the test image. It's just to make it easier to handle but you don't have to use them. I am currently not using them in any way and for training just generate random occlusion.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1754159": "how is everyone using all the masks available, there are some 4k+ mask, do we add all of them into the picture? or how are we doing it, wouldnt that require a lot of memory as every image would have 4k+ images to deal with", "1754308": "You should check out [How do you use the masks provided in the dataset?](https://www.kaggle.com/competitions/hotel-id-to-combat-human-trafficking-2022-fgvc9/discussion/313547) post. \nThe masks in train_masks folder are masks for test_images (the folder name is not correct so it's a little confusing).\n\nAbby Stylianou [Competition Host]:\n> The mask files are available for your convenience -- if there's a query image called 0001.jpg and a mask called 0001.png, the mask is simply a PNG that includes the exact mask from the query JPG. You can use these for processing of the query images in case it's easier than detecting the mask from the JPG, and also may also use any of the masks for whatever purpose in your training\n\nSo when doing predictions on the test dataset if there is image 0001.jpg and existing file with name 0001.png in the train_masks folder you can use the mask image to select the occlusion in the test image. It's just to make it easier to handle but you don't have to use them. I am currently not using them in any way and for training just generate random occlusion." }, "source": "meta" }