I am not an MD, but I have a masters in biology and I'm doing a PhD in oncology (bioinformatics side). I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labelling.
\n\nI edited this script for the natural conservancy fish so that I could label more than one object per image. I'm not sure that it's super helpful here but I kept it this way. My python script using OpenCV2 is also attached, use, edit or share as permitted by the licenses of sqshemet.
\n\nThe file is space separated with columns:
\n\nFilenames | number of rectangles | min(x1) | min(y1) | width1 | height1 (addtional coordinates if more than one rectangle).
\n\nx,y, width and height are pixel divided by two compared to the full size image (this was only due to the fact that some images did not fit my screen otherwise). Please take this into account when resizing.
\n\nI skipped some images when I had no idea about what should be annotated. I wanted to information as accurate as possible.
", "messages": [ { "id": "174995", "postDate": "04/13/2017 12:32:34", "content": "I am not an MD, but I have a masters in biology and I'm doing a PhD in oncology (bioinformatics side). I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labelling.
\n\nI edited this script for the natural conservancy fish so that I could label more than one object per image. I'm not sure that it's super helpful here but I kept it this way. My python script using OpenCV2 is also attached, use, edit or share as permitted by the licenses of sqshemet.
\n\nThe file is space separated with columns:
\n\nFilenames | number of rectangles | min(x1) | min(y1) | width1 | height1 (addtional coordinates if more than one rectangle).
\n\nx,y, width and height are pixel divided by two compared to the full size image (this was only due to the fact that some images did not fit my screen otherwise). Please take this into account when resizing.
\n\nI skipped some images when I had no idea about what should be annotated. I wanted to information as accurate as possible.
", "rawMarkdown": "# Disclaimer\n\nI am not an MD, but I have a masters in biology and I'm doing a PhD in oncology (bioinformatics side). I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labelling.\n\n# Script to annotate\nI edited this [script](http://web.mit.edu/sqshemet/Public/ObjectMarker.py) for the natural conservancy fish so that I could label more than one object per image. I'm not sure that it's super helpful here but I kept it this way. My python script using OpenCV2 is also attached, use, edit or share as permitted by the licenses of sqshemet.\n\n# Results\n## Format\nThe file is space separated with columns:\n\nFilenames | number of rectangles | min(x1) | min(y1) | width1 | height1 (addtional coordinates if more than one rectangle).\n\n## x,y ,width, height\nx,y, width and height are pixel divided by two compared to the full size image (this was only due to the fact that some images did not fit my screen otherwise). Please take this into account when resizing.\n\n##Additional notes:\n\n 1. The paths are \"Windows\" style, so you may want to change \"\\\" to \"/\" for Unix users\n 2. Feel free to comment or suggest improvements in the annotations.\n\n## Skipped images\nI skipped some images when I had no idea about what should be annotated. I wanted to information as accurate as possible.", "votes": null }, { "id": "175062", "postDate": "04/13/2017 16:44:35", "content": "A fellow Bioinformatician! :)
", "rawMarkdown": "A fellow Bioinformatician! :)", "votes": null }, { "id": "175064", "postDate": "04/13/2017 16:49:53", "content": "It can happen, even to the best of us! I hope that the boxes will be useful to some, and I didn't make too many mistakes (it would be a shame after saying that I did biology...)
", "rawMarkdown": "It can happen, even to the best of us! I hope that the boxes will be useful to some, and I didn't make too many mistakes (it would be a shame after saying that I did biology...)", "votes": null }, { "id": "175899", "postDate": "04/18/2017 02:59:21", "content": "Thanks Paul for the great efforts to come out the bounding box. Will you continue the annotations for the rest?
", "rawMarkdown": "Thanks Paul for the great efforts to come out the bounding box. Will you continue the annotations for the rest?", "votes": null }, { "id": "175919", "postDate": "04/18/2017 05:58:45", "content": "Hi,
\n\nI have just added bounding boxes for type 3 also. I don't plan on using and annotating the additional dataset so far if that was your question.
", "rawMarkdown": "Hi,\n\nI have just added bounding boxes for type 3 also. I don't plan on using and annotating the additional dataset so far if that was your question.", "votes": null }, { "id": "175959", "postDate": "04/18/2017 08:53:23", "content": "Hi Paul, type 3 is exactly what I asked for, thank you very much.
", "rawMarkdown": "Hi Paul, type 3 is exactly what I asked for, thank you very much.", "votes": null }, { "id": "175961", "postDate": "04/18/2017 08:57:08", "content": "Thanks Paul, type 3 is exactly what I asked for, really appreciated!
", "rawMarkdown": "Thanks Paul, type 3 is exactly what I asked for, really appreciated!", "votes": null }, { "id": "177172", "postDate": "04/24/2017 05:33:31", "content": "Hi Paul, Thanks for doing annotations. \nI have a question regarding x, y, width, and height values. You mentioned that these values are pixel divided by two, so in order to get actual bounding box on original image we have to multiply these values by 2 ?
", "rawMarkdown": "Hi Paul, Thanks for doing annotations. \nI have a question regarding x, y, width, and height values. You mentioned that these values are pixel divided by two, so in order to get actual bounding box on original image we have to multiply these values by 2 ?", "votes": null }, { "id": "177184", "postDate": "04/24/2017 06:45:01", "content": "Hi Neha,
\n\nI wanted to post a notebook to display the bounding boxes, but I cannot find how to import the annotation files.\nTo answer your question, the resize only impacts the window size, but not the output. So what you have are the original size.
", "rawMarkdown": "Hi Neha,\n\nI wanted to post a notebook to display the bounding boxes, but I cannot find how to import the annotation files.\nTo answer your question, the resize only impacts the window size, but not the output. So what you have are the original size.", "votes": null }, { "id": "177247", "postDate": "04/24/2017 12:11:05", "content": "thanks for your annoation!
", "rawMarkdown": "thanks for your annoation!", "votes": null }, { "id": "177468", "postDate": "04/24/2017 20:29:39", "content": "Just to let you know, I tried to upload the notebook I used to check boxes here: https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes\nI don't see anything so I don't know if the error is on my side or Kaggle or anything. Let me know if needed.
", "rawMarkdown": "Just to let you know, I tried to upload the notebook I used to check boxes here: [https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes][1]\nI don't see anything so I don't know if the error is on my side or Kaggle or anything. Let me know if needed.\n\n\n [1]: https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes", "votes": null }, { "id": "179050", "postDate": "04/29/2017 22:30:56", "content": "Seeking clarification:
\n\n\n\n\nI annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labeling.
\n
Epithelial lesions consist of squamous and columnar cells; what I'm seeing in these annotations for the most part seems like the bboxes are for the squamocolumnar junction, although occasionally they are for the entire cervical border. Can you confirm your original intent? Was it for the cervical border, the SJ or the OS?
\n\nThank you!
", "rawMarkdown": "Seeking clarification:\n\n> I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labeling.\n\nEpithelial lesions consist of squamous and columnar cells; what I'm seeing in these annotations for the _most_ part seems like the bboxes are for the squamocolumnar junction, although occasionally they are for the entire cervical border. Can you confirm your original intent? Was it for the cervical border, the SJ or the OS?\n\nThank you!", "votes": null }, { "id": "179135", "postDate": "04/30/2017 10:02:38", "content": "Hi,
\n\nI annotated everything that did not look like smooth epithelial tissue. I have to say that it's much harder than annotating animal images, since I'm much more familiar with the latter. I have annotated anything that seemed abnormal, for example presence of blood.
\n\nI you think that something has to be corrected, can you give an example (image number and what you would have annotated instead)?
", "rawMarkdown": "Hi,\n\nI annotated everything that did not look like smooth epithelial tissue. I have to say that it's much harder than annotating animal images, since I'm much more familiar with the latter. I have annotated anything that seemed abnormal, for example presence of blood.\n\nI you think that something has to be corrected, can you give an example (image number and what you would have annotated instead)?", "votes": null }, { "id": "186218", "postDate": "05/27/2017 04:12:03", "content": "Thank you for your annotation Paul.Much appreciated
", "rawMarkdown": "Thank you for your annotation Paul.Much appreciated", "votes": null }, { "id": "192344", "postDate": "06/13/2017 07:43:14", "content": "Thanks for your annoation. I have refined some annoations of yours and support for everybody. The file is in \"https://github.com/ChenhanXmu/Cervical_Cancer\". But I can't ensure that the work shall be done in the right way.
", "rawMarkdown": "Thanks for your annoation. I have refined some annoations of yours and support for everybody. The file is in \"https://github.com/ChenhanXmu/Cervical_Cancer\". But I can't ensure that the work shall be done in the right way.", "votes": null } ], "comments": [ { "id": 175062, "author_name": "craigglastonbury", "author_url": "", "post_date": "04/13/2017 16:44:35", "content": "A fellow Bioinformatician! :)
", "votes": null, "replies": [ { "id": 175064, "author_name": "deveaup", "author_url": "", "post_date": "04/13/2017 16:49:53", "content": "It can happen, even to the best of us! I hope that the boxes will be useful to some, and I didn't make too many mistakes (it would be a shame after saying that I did biology...)
", "votes": null, "replies": [] } ] }, { "id": 175899, "author_name": "yangxulei", "author_url": "", "post_date": "04/18/2017 02:59:21", "content": "Thanks Paul for the great efforts to come out the bounding box. Will you continue the annotations for the rest?
", "votes": null, "replies": [ { "id": 175919, "author_name": "deveaup", "author_url": "", "post_date": "04/18/2017 05:58:45", "content": "Hi,
\n\nI have just added bounding boxes for type 3 also. I don't plan on using and annotating the additional dataset so far if that was your question.
", "votes": null, "replies": [] }, { "id": 175959, "author_name": "yangxulei", "author_url": "", "post_date": "04/18/2017 08:53:23", "content": "Hi Paul, type 3 is exactly what I asked for, thank you very much.
", "votes": null, "replies": [] }, { "id": 175961, "author_name": "yangxulei", "author_url": "", "post_date": "04/18/2017 08:57:08", "content": "Thanks Paul, type 3 is exactly what I asked for, really appreciated!
", "votes": null, "replies": [] } ] }, { "id": 177172, "author_name": "nehaboob", "author_url": "", "post_date": "04/24/2017 05:33:31", "content": "Hi Paul, Thanks for doing annotations. \nI have a question regarding x, y, width, and height values. You mentioned that these values are pixel divided by two, so in order to get actual bounding box on original image we have to multiply these values by 2 ?
", "votes": null, "replies": [ { "id": 177184, "author_name": "deveaup", "author_url": "", "post_date": "04/24/2017 06:45:01", "content": "Hi Neha,
\n\nI wanted to post a notebook to display the bounding boxes, but I cannot find how to import the annotation files.\nTo answer your question, the resize only impacts the window size, but not the output. So what you have are the original size.
", "votes": null, "replies": [] }, { "id": 177468, "author_name": "deveaup", "author_url": "", "post_date": "04/24/2017 20:29:39", "content": "Just to let you know, I tried to upload the notebook I used to check boxes here: https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes\nI don't see anything so I don't know if the error is on my side or Kaggle or anything. Let me know if needed.
", "votes": null, "replies": [] } ] }, { "id": 177247, "author_name": "chen1993", "author_url": "", "post_date": "04/24/2017 12:11:05", "content": "thanks for your annoation!
", "votes": null, "replies": [] }, { "id": 179050, "author_name": "authman", "author_url": "", "post_date": "04/29/2017 22:30:56", "content": "Seeking clarification:
\n\n\n\n\nI annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labeling.
\n
Epithelial lesions consist of squamous and columnar cells; what I'm seeing in these annotations for the most part seems like the bboxes are for the squamocolumnar junction, although occasionally they are for the entire cervical border. Can you confirm your original intent? Was it for the cervical border, the SJ or the OS?
\n\nThank you!
", "votes": null, "replies": [ { "id": 179135, "author_name": "deveaup", "author_url": "", "post_date": "04/30/2017 10:02:38", "content": "Hi,
\n\nI annotated everything that did not look like smooth epithelial tissue. I have to say that it's much harder than annotating animal images, since I'm much more familiar with the latter. I have annotated anything that seemed abnormal, for example presence of blood.
\n\nI you think that something has to be corrected, can you give an example (image number and what you would have annotated instead)?
", "votes": null, "replies": [] } ] }, { "id": 186218, "author_name": "rteja1113", "author_url": "", "post_date": "05/27/2017 04:12:03", "content": "Thank you for your annotation Paul.Much appreciated
", "votes": null, "replies": [] }, { "id": 192344, "author_name": "chenhanxmu199214", "author_url": "", "post_date": "06/13/2017 07:43:14", "content": "Thanks for your annoation. I have refined some annoations of yours and support for everybody. The file is in \"https://github.com/ChenhanXmu/Cervical_Cancer\". But I can't ensure that the work shall be done in the right way.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "174995": "# Disclaimer\n\nI am not an MD, but I have a masters in biology and I'm doing a PhD in oncology (bioinformatics side). I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labelling.\n\n# Script to annotate\nI edited this [script](http://web.mit.edu/sqshemet/Public/ObjectMarker.py) for the natural conservancy fish so that I could label more than one object per image. I'm not sure that it's super helpful here but I kept it this way. My python script using OpenCV2 is also attached, use, edit or share as permitted by the licenses of sqshemet.\n\n# Results\n## Format\nThe file is space separated with columns:\n\nFilenames | number of rectangles | min(x1) | min(y1) | width1 | height1 (addtional coordinates if more than one rectangle).\n\n## x,y ,width, height\nx,y, width and height are pixel divided by two compared to the full size image (this was only due to the fact that some images did not fit my screen otherwise). Please take this into account when resizing.\n\n##Additional notes:\n\n 1. The paths are \"Windows\" style, so you may want to change \"\\\" to \"/\" for Unix users\n 2. Feel free to comment or suggest improvements in the annotations.\n\n## Skipped images\nI skipped some images when I had no idea about what should be annotated. I wanted to information as accurate as possible.", "175062": "A fellow Bioinformatician! :)", "175064": "It can happen, even to the best of us! I hope that the boxes will be useful to some, and I didn't make too many mistakes (it would be a shame after saying that I did biology...)", "175899": "Thanks Paul for the great efforts to come out the bounding box. Will you continue the annotations for the rest?", "175919": "Hi,\n\nI have just added bounding boxes for type 3 also. I don't plan on using and annotating the additional dataset so far if that was your question.", "175959": "Hi Paul, type 3 is exactly what I asked for, thank you very much.", "175961": "Thanks Paul, type 3 is exactly what I asked for, really appreciated!", "177172": "Hi Paul, Thanks for doing annotations. \nI have a question regarding x, y, width, and height values. You mentioned that these values are pixel divided by two, so in order to get actual bounding box on original image we have to multiply these values by 2 ?", "177184": "Hi Neha,\n\nI wanted to post a notebook to display the bounding boxes, but I cannot find how to import the annotation files.\nTo answer your question, the resize only impacts the window size, but not the output. So what you have are the original size.", "177247": "thanks for your annoation!", "177468": "Just to let you know, I tried to upload the notebook I used to check boxes here: [https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes][1]\nI don't see anything so I don't know if the error is on my side or Kaggle or anything. Let me know if needed.\n\n\n [1]: https://www.kaggle.com/deveaup/intel-mobileodt-cervical-cancer-screening/checking-bounding-boxes", "179050": "Seeking clarification:\n\n> I annotated on all images what looked like epithelial lesions, but I cannot guarantee perfect labeling.\n\nEpithelial lesions consist of squamous and columnar cells; what I'm seeing in these annotations for the _most_ part seems like the bboxes are for the squamocolumnar junction, although occasionally they are for the entire cervical border. Can you confirm your original intent? Was it for the cervical border, the SJ or the OS?\n\nThank you!", "179135": "Hi,\n\nI annotated everything that did not look like smooth epithelial tissue. I have to say that it's much harder than annotating animal images, since I'm much more familiar with the latter. I have annotated anything that seemed abnormal, for example presence of blood.\n\nI you think that something has to be corrected, can you give an example (image number and what you would have annotated instead)?", "186218": "Thank you for your annotation Paul.Much appreciated", "192344": "Thanks for your annoation. I have refined some annoations of yours and support for everybody. The file is in \"https://github.com/ChenhanXmu/Cervical_Cancer\". But I can't ensure that the work shall be done in the right way." }, "source": "meta" }