{ "id": 186730, "title": "Network Extraction from Images", "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/186730", "author_name": "Pythonian", "post_date": "2020-09-25T16:42:57.359000", "votes": 0, "comment_count": 0, "views": 0, "content": "

To be honest, this post is a knockoff of my other post in the Covid19 mRNA competition, but I think that this cross between fields (graph theory & images), could provide some useful results. So far, the images in this competition have not provided a huge boost considering the best metric, as the current best shared kernel uses these images, and scores -6.807, while the best possible score is ~-4.595. I propose that we try out automatic graph network extraction on the Osic images, using the module nefi. You should be able to both enhance the graphs and compute statistics on them with networkx.
\nNefi:

\n", "messages": [ { "id": 1026924, "postDate": "2020-09-25T16:42:57.360Z", "content": "

To be honest, this post is a knockoff of my other post in the Covid19 mRNA competition, but I think that this cross between fields (graph theory & images), could provide some useful results. So far, the images in this competition have not provided a huge boost considering the best metric, as the current best shared kernel uses these images, and scores -6.807, while the best possible score is ~-4.595. I propose that we try out automatic graph network extraction on the Osic images, using the module nefi. You should be able to both enhance the graphs and compute statistics on them with networkx.
\nNefi:

\n", "rawMarkdown": "To be honest, this post is a knockoff of my other [post ](https://www.kaggle.com/c/stanford-covid-vaccine/discussion/186723) in the [Covid19 mRNA competition](https://www.kaggle.com/c/stanford-covid-vaccine/overview), but I think that this cross between fields (graph theory & images), could provide some useful results. So far, the images in this competition have not provided a huge boost considering the best metric, as the current [best shared kernel](https://www.kaggle.com/thebigd8ta/higher-lb-score-by-tuning-mloss-upgrade-1696e2) uses these images, and scores -6.807, while the best possible score is ~-4.595. I propose that we try out automatic graph network extraction on the Osic images, using the module nefi. You should be able to both enhance the graphs and compute statistics on them with networkx.\n**Nefi:**\n\n- https://nefi.mpi-inf.mpg.de/documentation/index.html\n- https://nefi.mpi-inf.mpg.de/guide.php#contribution\n- https://nefi.mpi-inf.mpg.de/documentation/rst_files/Quick_Start_Guide_for_users.html", "votes": 1 } ], "comments": [], "raw_markdown_by_id": { "1026924": "To be honest, this post is a knockoff of my other [post ](https://www.kaggle.com/c/stanford-covid-vaccine/discussion/186723) in the [Covid19 mRNA competition](https://www.kaggle.com/c/stanford-covid-vaccine/overview), but I think that this cross between fields (graph theory & images), could provide some useful results. So far, the images in this competition have not provided a huge boost considering the best metric, as the current [best shared kernel](https://www.kaggle.com/thebigd8ta/higher-lb-score-by-tuning-mloss-upgrade-1696e2) uses these images, and scores -6.807, while the best possible score is ~-4.595. I propose that we try out automatic graph network extraction on the Osic images, using the module nefi. You should be able to both enhance the graphs and compute statistics on them with networkx.\n**Nefi:**\n\n- https://nefi.mpi-inf.mpg.de/documentation/index.html\n- https://nefi.mpi-inf.mpg.de/guide.php#contribution\n- https://nefi.mpi-inf.mpg.de/documentation/rst_files/Quick_Start_Guide_for_users.html" } }