Not sure if it has already been posted, but if you need data for the bigger amount of roads, then here you go:
\nhttps://www.kaggle.com/google/android-smartphones-high-accuracy-datasets
", "messages": [ { "id": "1377078", "postDate": "07/05/2021 15:09:14", "content": "Not sure if it has already been posted, but if you need data for the bigger amount of roads, then here you go:
\nhttps://www.kaggle.com/google/android-smartphones-high-accuracy-datasets
", "rawMarkdown": "Not sure if it has already been posted, but if you need data for the bigger amount of roads, then here you go:\n\nhttps://www.kaggle.com/google/android-smartphones-high-accuracy-datasets", "votes": null }, { "id": "1399205", "postDate": "07/25/2021 02:39:01", "content": "Thanks for sharing. These look to be the same collections as in the competition dataset- is there something I'm missing?
", "rawMarkdown": "Thanks for sharing. These look to be the same collections as in the competition dataset- is there something I'm missing?", "votes": null }, { "id": "1399403", "postDate": "07/25/2021 09:24:29", "content": "From what I see for some of the pathes at least start/end points are different. I will post a notebook with explanation how to use those files though.
", "rawMarkdown": "From what I see for some of the pathes at least start/end points are different. I will post a notebook with explanation how to use those files though.", "votes": null }, { "id": "1399539", "postDate": "07/25/2021 12:21:41", "content": "@robikscube
\n![\"\"](\"https://i.postimg.cc/KYSxdHNP/compim.png\")
\nWell, while holding the same name and mainly following the same path, parsing the lat/lng truth directly from the .nmea files gives a bit different results. Here you can the the results for some random area and yellow points are the ones manually parsed from the .nmeas in different dataset. I dont really think that this is a rounding error, because one can clearly see that gap exists for blue points but not for the yellow ones. And vice versa, yellow dots sometimes behave very strange.
Here is the parsing code, so I dont really know how else can I comment on that or if I am missing something.
\nhttps://www.kaggle.com/avtobusbratiev/additional-gt-data
Thanks @avtobusbratiev for the response. This is interesting, I'll have to look into this further if I have time. I noticed your notebook uses the *chipset.nmea files, but the data description says the ground truth data is found in these files:
SPAN_{Phone name}_{Report rate}.nmea: Ground truth positions at the phone's position
I have just noticed that. Yes, this is my mistake. Have to add the 'chipset location' in the beginning.
\nStill, there are different SPAN files and 1 Hz/10 Hz gt is pretty different in terms of precise lat/lng, while official training data uses only 10Hz measurements. I will look more into it, maybe chipset is of any use after all.
", "rawMarkdown": "I have just noticed that. Yes, this is my mistake. Have to add the 'chipset location' in the beginning. \n\nStill, there are different SPAN files and 1 Hz/10 Hz gt is pretty different in terms of precise lat/lng, while official training data uses only 10Hz measurements. I will look more into it, maybe chipset is of any use after all.", "votes": null } ], "comments": [ { "id": 1399205, "author_name": "robikscube", "author_url": "", "post_date": "07/25/2021 02:39:01", "content": "Thanks for sharing. These look to be the same collections as in the competition dataset- is there something I'm missing?
", "votes": null, "replies": [ { "id": 1399403, "author_name": "avtobusbratiev", "author_url": "", "post_date": "07/25/2021 09:24:29", "content": "From what I see for some of the pathes at least start/end points are different. I will post a notebook with explanation how to use those files though.
", "votes": null, "replies": [] }, { "id": 1399539, "author_name": "avtobusbratiev", "author_url": "", "post_date": "07/25/2021 12:21:41", "content": "@robikscube
\n
\nWell, while holding the same name and mainly following the same path, parsing the lat/lng truth directly from the .nmea files gives a bit different results. Here you can the the results for some random area and yellow points are the ones manually parsed from the .nmeas in different dataset. I dont really think that this is a rounding error, because one can clearly see that gap exists for blue points but not for the yellow ones. And vice versa, yellow dots sometimes behave very strange.
Here is the parsing code, so I dont really know how else can I comment on that or if I am missing something.
\nhttps://www.kaggle.com/avtobusbratiev/additional-gt-data
Thanks @avtobusbratiev for the response. This is interesting, I'll have to look into this further if I have time. I noticed your notebook uses the *chipset.nmea files, but the data description says the ground truth data is found in these files:
SPAN_{Phone name}_{Report rate}.nmea: Ground truth positions at the phone's position
I have just noticed that. Yes, this is my mistake. Have to add the 'chipset location' in the beginning.
\nStill, there are different SPAN files and 1 Hz/10 Hz gt is pretty different in terms of precise lat/lng, while official training data uses only 10Hz measurements. I will look more into it, maybe chipset is of any use after all.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1377078": "Not sure if it has already been posted, but if you need data for the bigger amount of roads, then here you go:\n\nhttps://www.kaggle.com/google/android-smartphones-high-accuracy-datasets", "1399205": "Thanks for sharing. These look to be the same collections as in the competition dataset- is there something I'm missing?", "1399403": "From what I see for some of the pathes at least start/end points are different. I will post a notebook with explanation how to use those files though.", "1399539": "robikscube \n\nWell, while holding the same name and mainly following the same path, parsing the lat/lng truth directly from the .nmea files gives a bit different results. Here you can the the results for some random area and yellow points are the ones manually parsed from the .nmeas in different dataset. I dont really think that this is a rounding error, because one can clearly see that gap exists for blue points but not for the yellow ones. And vice versa, yellow dots sometimes behave very strange.\n\nHere is the parsing code, so I dont really know how else can I comment on that or if I am missing something.\nhttps://www.kaggle.com/avtobusbratiev/additional-gt-data", "1399604": "Thanks @avtobusbratiev for the response. This is interesting, I'll have to look into this further if I have time. I noticed your notebook uses the `*chipset.nmea` files, but the data description says the ground truth data is found in these files:\n\n `SPAN_{Phone name}_{Report rate}.nmea: Ground truth positions at the phone's position`", "1399647": "I have just noticed that. Yes, this is my mistake. Have to add the 'chipset location' in the beginning. \n\nStill, there are different SPAN files and 1 Hz/10 Hz gt is pretty different in terms of precise lat/lng, while official training data uses only 10Hz measurements. I will look more into it, maybe chipset is of any use after all." }, "source": "meta" }