I was wondering if it is useful to train the model for 3 classes (background, blood vessel and glomerulus) or if it was sufficient to not include glomeruli mask and prediction class in the training of the model. Since the aim of this challenge is to identify blood vessels and not glomeruli, do we need this additional information?
", "messages": [ { "id": 2286894, "postDate": "2023-06-03T23:09:19.743Z", "content": "I was wondering if it is useful to train the model for 3 classes (background, blood vessel and glomerulus) or if it was sufficient to not include glomeruli mask and prediction class in the training of the model. Since the aim of this challenge is to identify blood vessels and not glomeruli, do we need this additional information?
", "rawMarkdown": "I was wondering if it is useful to train the model for 3 classes (background, blood vessel and glomerulus) or if it was sufficient to not include glomeruli mask and prediction class in the training of the model. Since the aim of this challenge is to identify blood vessels and not glomeruli, do we need this additional information?", "votes": 19 }, { "id": 2290366, "postDate": "2023-06-06T18:10:34.250Z", "content": "Hello,
\nThe annotations for glomeruli in the test set will be available during submission. You may use these to exclude your blood vessel predictions that lie within the glomeruli structures. You don't need to predict glomeruli.
I am aware of polygons.jsonl where the annotations for the train set resides. Where can we get these annotations for glomerulus during inference in order to use these in excluding the blood vessel predictions within them.
", "rawMarkdown": "I am aware of polygons.jsonl where the annotations for the train set resides. **Where** can we get these annotations for glomerulus during **inference** in order to use these in excluding the blood vessel predictions within them. ", "votes": 7, "replies": [ { "id": 2295036, "postDate": "2023-06-10T14:28:29.790Z", "content": "Hi @alabibojesomo, the hidden version of the polygons.jsonl file should contain glomerulus annotations for the test set images. They aren't present in the public version of the dataset, but are added in the hidden version available to your notebook during inference. (I just double checked that they are indeed there.)
Note that not all test set images will have their id present in polygons.jsonl; only those with glomerulus structures will be present.
Thank you so much, this is a very valuable clarification!
", "rawMarkdown": "Thank you so much, this is a very valuable clarification!" }, { "id": 2330152, "postDate": "2023-07-04T18:44:50.163Z", "content": "@ryanholbrook @yashvrdnjain Can you specify the location of the polygons.jsonl and how it is structured? When I try to include the polygons.jsonl in my code during submission I only get the message the the notebook with the test data threw an exception while the one with the fake test data runs without issues.
\nAre there any keys that do not exist or are renamed such as \"id\", \"annotations\", \"type\" or \"coordinates\"?
there is only image_id.tif under test folder, where to find its glomeruli annotations?
", "rawMarkdown": "there is only image_id.tif under test folder, where to find its glomeruli annotations?" }, { "id": 2293454, "postDate": "2023-06-09T07:44:25.730Z", "content": "Hello,
\nIt appears that polygons.jsonl does not contain the image id from the test folder, neither during submit nor during training. Therefore, I join the question. Where can we get such information?
Will the annotations for unsure be included as well. If not, how will these regions be handled?
", "rawMarkdown": "Will the annotations for unsure be included as well. If not, how will these regions be handled?" } ] }, { "id": 2288215, "postDate": "2023-06-05T08:43:04.047Z", "content": "Maybe this can be useful to avoid the blood_vessel prediction results that are in the glomerulus. Because if this is loaded, it will be counted as a false positive
\n\n", "rawMarkdown": "Maybe this can be useful to avoid the blood_vessel prediction results that are in the glomerulus. Because if this is loaded, it will be counted as a false positive\n>glomerulus A capillary ball structure in the kidney. These parts of the images were excluded from blood vessel annotation. **You should ensure none of your test set predictions occur within glomerulus structures as they will be counted as false positives.** Annotations are provided for test set tiles.", "votes": 4, "replies": [ { "id": 2288635, "postDate": "2023-06-05T14:33:50.923Z", "content": "glomerulus A capillary ball structure in the kidney. These parts of the images were excluded from blood vessel annotation. You should ensure none of your test set predictions occur within glomerulus structures as they will be counted as false positives. Annotations are provided for test set tiles.
\n
I overread this paragraph! Thank you! Further, the state that the \"test tiles contain annotations for the glomeruli\". Does this mean that in tje submission, we have access to the polygon coordinates of glomeruli and can then exclude predictions for blood vessels from these regions?
", "rawMarkdown": "I overread this paragraph! Thank you! Further, the state that the \"test tiles contain annotations for the glomeruli\". Does this mean that in tje submission, we have access to the polygon coordinates of glomeruli and can then exclude predictions for blood vessels from these regions?", "replies": [ { "id": 2289196, "postDate": "2023-06-06T02:17:03.800Z", "content": "No, if needed, you will have to predict the glomerulus as well and used tour glomerulusbpredictions to decide on which blood vessels are to be excluded based on being within your predicted glomwrulus instances.
", "rawMarkdown": "No, if needed, you will have to predict the glomerulus as well and used tour glomerulusbpredictions to decide on which blood vessels are to be excluded based on being within your predicted glomwrulus instances.", "votes": -5 } ] }, { "id": 2355507, "postDate": "2023-07-23T11:58:50.440Z", "content": "What do we exactly mean by \"These parts of the images were excluded from blood vessel annotation. \" ?
", "rawMarkdown": "What do we exactly mean by **\"These parts of the images were excluded from blood vessel annotation. \"** ?", "replies": [ { "id": 2367497, "postDate": "2023-07-31T15:29:42.690Z", "content": "It will ending up with a drop in LB/PB score if some of the objects are touching each other, especially for glomeruli and blood vessels.
", "rawMarkdown": "It will ending up with a drop in LB/PB score if some of the objects are touching each other, especially for glomeruli and blood vessels.", "isDeleted": true } ] } ] }, { "id": 2365158, "postDate": "2023-07-30T03:21:40.730Z", "content": "Can someone explain what this is talking about? How do I add that to this notebook here: https://www.kaggle.com/code/nghiahoangtrung/hubmap-mmdet3-1-single-fold-inference ?
\nThanks.
", "rawMarkdown": "Can someone explain what this is talking about? How do I add that to this notebook here: https://www.kaggle.com/code/nghiahoangtrung/hubmap-mmdet3-1-single-fold-inference ?\n\nThanks.", "votes": -1 } ], "comments": [ { "id": 2290366, "author_name": "Yashvardhan Jain", "author_url": "", "post_date": "2023-06-06T18:10:34.250000", "content": "Hello,
\nThe annotations for glomeruli in the test set will be available during submission. You may use these to exclude your blood vessel predictions that lie within the glomeruli structures. You don't need to predict glomeruli.
I am aware of polygons.jsonl where the annotations for the train set resides. Where can we get these annotations for glomerulus during inference in order to use these in excluding the blood vessel predictions within them.
", "votes": 7, "replies": [ { "id": 2295036, "author_name": "Ryan Holbrook", "author_url": "", "post_date": "2023-06-10T14:28:29.790000", "content": "Hi @alabibojesomo, the hidden version of the polygons.jsonl file should contain glomerulus annotations for the test set images. They aren't present in the public version of the dataset, but are added in the hidden version available to your notebook during inference. (I just double checked that they are indeed there.)
Note that not all test set images will have their id present in polygons.jsonl; only those with glomerulus structures will be present.
Thank you so much, this is a very valuable clarification!
", "votes": 0, "replies": [] }, { "id": 2330152, "author_name": "John Noman", "author_url": "", "post_date": "2023-07-04T18:44:50.163000", "content": "@ryanholbrook @yashvrdnjain Can you specify the location of the polygons.jsonl and how it is structured? When I try to include the polygons.jsonl in my code during submission I only get the message the the notebook with the test data threw an exception while the one with the fake test data runs without issues.
\nAre there any keys that do not exist or are renamed such as \"id\", \"annotations\", \"type\" or \"coordinates\"?
there is only image_id.tif under test folder, where to find its glomeruli annotations?
", "votes": 0, "replies": [] }, { "id": 2293454, "author_name": "tsobolev", "author_url": "", "post_date": "2023-06-09T07:44:25.730000", "content": "Hello,
\nIt appears that polygons.jsonl does not contain the image id from the test folder, neither during submit nor during training. Therefore, I join the question. Where can we get such information?
Will the annotations for unsure be included as well. If not, how will these regions be handled?
", "votes": 0, "replies": [] } ] }, { "id": 2288215, "author_name": "Harits Fadlilah", "author_url": "", "post_date": "2023-06-05T08:43:04.047000", "content": "Maybe this can be useful to avoid the blood_vessel prediction results that are in the glomerulus. Because if this is loaded, it will be counted as a false positive
\n\n", "votes": 4, "replies": [ { "id": 2288635, "author_name": "Jan H", "author_url": "", "post_date": "2023-06-05T14:33:50.923000", "content": "glomerulus A capillary ball structure in the kidney. These parts of the images were excluded from blood vessel annotation. You should ensure none of your test set predictions occur within glomerulus structures as they will be counted as false positives. Annotations are provided for test set tiles.
\n
I overread this paragraph! Thank you! Further, the state that the \"test tiles contain annotations for the glomeruli\". Does this mean that in tje submission, we have access to the polygon coordinates of glomeruli and can then exclude predictions for blood vessels from these regions?
", "votes": 0, "replies": [ { "id": 2289196, "author_name": "HungryLearner", "author_url": "", "post_date": "2023-06-06T02:17:03.800000", "content": "No, if needed, you will have to predict the glomerulus as well and used tour glomerulusbpredictions to decide on which blood vessels are to be excluded based on being within your predicted glomwrulus instances.
", "votes": -5, "replies": [] } ] }, { "id": 2355507, "author_name": "Kethupio", "author_url": "", "post_date": "2023-07-23T11:58:50.440000", "content": "What do we exactly mean by \"These parts of the images were excluded from blood vessel annotation. \" ?
", "votes": 0, "replies": [ { "id": 2367497, "author_name": "", "author_url": "", "post_date": "2023-07-31T15:29:42.690000", "content": "It will ending up with a drop in LB/PB score if some of the objects are touching each other, especially for glomeruli and blood vessels.
", "votes": 0, "replies": [] } ] } ] }, { "id": 2365158, "author_name": "Bo Peng", "author_url": "", "post_date": "2023-07-30T03:21:40.730000", "content": "Can someone explain what this is talking about? How do I add that to this notebook here: https://www.kaggle.com/code/nghiahoangtrung/hubmap-mmdet3-1-single-fold-inference ?
\nThanks.
", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "2286894": "I was wondering if it is useful to train the model for 3 classes (background, blood vessel and glomerulus) or if it was sufficient to not include glomeruli mask and prediction class in the training of the model. Since the aim of this challenge is to identify blood vessels and not glomeruli, do we need this additional information?", "2290366": "Hello, \nThe annotations for glomeruli in the test set will be available during submission. You may use these to exclude your blood vessel predictions that lie within the glomeruli structures. You don't need to predict glomeruli. ", "2288215": "Maybe this can be useful to avoid the blood_vessel prediction results that are in the glomerulus. Because if this is loaded, it will be counted as a false positive\n>glomerulus A capillary ball structure in the kidney. These parts of the images were excluded from blood vessel annotation. **You should ensure none of your test set predictions occur within glomerulus structures as they will be counted as false positives.** Annotations are provided for test set tiles.", "2365158": "Can someone explain what this is talking about? How do I add that to this notebook here: https://www.kaggle.com/code/nghiahoangtrung/hubmap-mmdet3-1-single-fold-inference ?\n\nThanks." } }