{ "id": 583676, "title": "What is the sample rate on the IMU device?", "url": "/competitions/cmi-detect-behavior-with-sensor-data/discussion/583676", "author_name": "", "post_date": "2025-06-08T13:26:23.501460Z", "votes": 5, "comment_count": 8, "views": 0, "content": "

Hi,

\n

What is the sample rate on the IMU device?

", "messages": [ { "id": "3219904", "postDate": "06/08/2025 13:26:23", "content": "

Hi,

\n

What is the sample rate on the IMU device?

", "rawMarkdown": "Hi, \n\nWhat is the sample rate on the IMU device?", "votes": null }, { "id": "3219906", "postDate": "06/08/2025 13:38:48", "content": "

TL;DR: Given the column names, we’re almost certainly looking at fused outputs. The BNO080/BNO085 doesn’t run every channel at a fixed 1 kHz. You request a report rate R, and it picks the nearest supported ODR within

\n
.·R ≤ ConfiguredRate ≤ .·R\n
\n

up to each feature’s ceiling (e.g. raw accel 500 Hz, raw gyro 400 Hz, gyro-rotation vector 1 000 Hz, etc.). You enable whichever raw or fused streams you need, and the sensor hub programs its MEMS ODRs and MotionEngine fusion core accordingly.

\n
\n

First Comment

\n

Hi,
\nFrom https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/overview:

\n
\n

1x Inertial Measurement Unit (IMU; BNO080/BNO085): An integrated sensor that combines accelerometer, gyroscope, and magnetometer measurements with onboard processing to provide orientation and motion data.

\n
\n

The Helios IMU (BNO080/BNO085 SiP) seems to deliver a fixed 1 kHz sampling rate. In the product brief you’ll find:

\n
\n

“1 KHz Sample Rate – Enables flawless, smooth head-tracking with low latency…”

\n
\n

No alternate rates are listed, so you might consider the sample rate to be fixed at 1 kHz.

\n
\n

Update 1

\n

The competition linked pdf calls out the Helios IMU (BNO080/BNO085) as having a “1 kHz Sample Rate – Enables flawless, smooth head-tracking with low latency…” (from the product brief), which might suggest a fixed data rate.
\nHowever, the official BNO08X datasheet shows that all sensor and fusion outputs are configurable by the requested report rate R, and that the device picks the nearest supported Output Data Rate (ODR) such that

\n
. · R ≤ ConfiguredRate ≤ . · R\n
\n

Each feature has its own maximum ceiling:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Feature / SensorMax. Report RateFused?
Raw accelerometer500 HzNo
Raw gyroscope400 HzNo
Raw magnetometer100 HzNo
Gyro-rotation vector1 000 HzYes (gyro+accel+mag)
Rotation vector400 HzYes (accel+gyro+mag)
Gaming rotation vector400 HzYes (accel+gyro)
Geomagnetic rotation vector90 HzYes (accel+mag)
Gravity400 HzYes (accel+gyro)
Linear acceleration400 HzYes (accel+gyro)
\n

So rather than a hard-wired 1 kHz on all channels, you request whatever rate you need (up to each feature’s maximum) and the BNO08X will program the underlying MEMS ODRs and/or MotionEngine fusion core accordingly.

\n

—

\n

*Source: BNO080/085 Datasheet: https://www.ceva-ip.com/wp-content/uploads/BNO080_085-Datasheet.pdf*

", "rawMarkdown": "**TL;DR:** Given the column names, we’re almost certainly looking at fused outputs. The BNO080/BNO085 doesn’t run every channel at a fixed 1 kHz. You request a report rate *R*, and it picks the nearest supported ODR within\n\n```\n0.9·R ≤ ConfiguredRate ≤ 2.1·R\n```\n\nup to each feature’s ceiling (e.g. raw accel 500 Hz, raw gyro 400 Hz, gyro-rotation vector 1 000 Hz, etc.). You enable whichever raw or fused streams you need, and the sensor hub programs its MEMS ODRs and MotionEngine fusion core accordingly.\n\n---\n\n## First Comment\nHi,\nFrom https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/overview:\n>1x Inertial Measurement Unit (IMU; BNO080/BNO085): An integrated sensor that combines accelerometer, gyroscope, and magnetometer measurements with onboard processing to provide orientation and motion data.\n\nThe Helios IMU (BNO080/BNO085 SiP) seems to deliver a fixed 1 kHz sampling rate. In the product brief you’ll find:\n> “1 KHz Sample Rate – Enables flawless, smooth head-tracking with low latency…”\n\nNo alternate rates are listed, so you might consider the sample rate to be fixed at 1 kHz.\n\n---\n## Update 1\nThe competition linked pdf calls out the Helios IMU (BNO080/BNO085) as having a “1 kHz Sample Rate – Enables flawless, smooth head-tracking with low latency…” (from the product brief), which might suggest a fixed data rate.\nHowever, the official BNO08X datasheet shows that **all** sensor and fusion outputs are *configurable* by the requested report rate R, and that the device picks the nearest supported Output Data Rate (ODR) such that\n\n```\n0.9 · R ≤ ConfiguredRate ≤ 2.1 · R\n```\n\nEach feature has its own maximum ceiling:\n\n| Feature / Sensor | Max. Report Rate | Fused? |\n| ------------------------------- | :--------------: | :------------------: |\n| **Raw accelerometer** | 500 Hz | No |\n| **Raw gyroscope** | 400 Hz | No |\n| **Raw magnetometer** | 100 Hz | No |\n| **Gyro-rotation vector** | 1 000 Hz | Yes (gyro+accel+mag) |\n| **Rotation vector** | 400 Hz | Yes (accel+gyro+mag) |\n| **Gaming rotation vector** | 400 Hz | Yes (accel+gyro) |\n| **Geomagnetic rotation vector** | 90 Hz | Yes (accel+mag) |\n| **Gravity** | 400 Hz | Yes (accel+gyro) |\n| **Linear acceleration** | 400 Hz | Yes (accel+gyro) |\n\nSo rather than a hard-wired 1 kHz on *all* channels, you request whatever rate you need (up to each feature’s maximum) and the BNO08X will program the underlying MEMS ODRs and/or MotionEngine fusion core accordingly.\n\n—\n\n*Source: BNO080/085 Datasheet: https://www.ceva-ip.com/wp-content/uploads/BNO080_085-Datasheet.pdf*", "votes": null }, { "id": "3219930", "postDate": "06/08/2025 14:26:55", "content": "

This would imply we have to classify the position based on observations that last ~0.1 second each

", "rawMarkdown": "This would imply we have to classify the position based on observations that last ~0.1 second each", "votes": null }, { "id": "3219972", "postDate": "06/08/2025 15:27:42", "content": "

@defdet Exacly what I thought… just doesn't add up.
\nLooks 50Hz'ish by the provided videos (about 1.5 second movements).
\nHope competition organizers will shed some light.

", "rawMarkdown": "defdet Exacly what I thought... just doesn't add up.\nLooks 50Hz'ish by the provided videos (about 1.5 second movements).\nHope competition organizers will shed some light.", "votes": null }, { "id": "3219979", "postDate": "06/08/2025 15:41:24", "content": "

@roysegalz @defdet
\nUpdated my comment.
\nhttps://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583676#3219906

", "rawMarkdown": "roysegalz @defdet \nUpdated my comment. \nhttps://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583676#3219906", "votes": null }, { "id": "3220013", "postDate": "06/08/2025 16:31:19", "content": "

yeah, it would be nice to receive some concrete answer, instead of having to speculate

", "rawMarkdown": "yeah, it would be nice to receive some concrete answer, instead of having to speculate", "votes": null }, { "id": "3220028", "postDate": "06/08/2025 17:06:21", "content": "

Well absolutely, I did not know how this IMU works ;) Sorry for bothering you.

", "rawMarkdown": "Well absolutely, I did not know how this IMU works ;) Sorry for bothering you.", "votes": null }, { "id": "3220226", "postDate": "06/09/2025 03:18:25", "content": "

as per the visualizer script, shared here: https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583118

\n
 ():\n    ...\n\n ():\n    ...\n    parser.add_argument(\n        , =, default=, =\n    )\n    ...\n
\n

the data can be interpreted at a sampling rate of 10 Hz (10 rows per second)

", "rawMarkdown": "as per the visualizer script, shared here: https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583118\n\n```python\n\ndef render_animation(sequence_data, output_file, framerate=10):\n ...\n\ndef main():\n ...\n parser.add_argument(\n \"--framerate\", type=int, default=10, help=\"Frame rate for output video\"\n )\n ...\n```\n\nthe data can be interpreted at a sampling rate of 10 Hz (10 rows per second)", "votes": null }, { "id": "3227525", "postDate": "06/19/2025 04:26:58", "content": "

It looks reasonable, also it's completely in line with demo videos from competition page.

", "rawMarkdown": "It looks reasonable, also it's completely in line with demo videos from competition page.", "votes": null } ], "comments": [ { "id": 3219906, "author_name": "maxstickel", "author_url": "", "post_date": "06/08/2025 13:38:48", "content": "

TL;DR: Given the column names, we’re almost certainly looking at fused outputs. The BNO080/BNO085 doesn’t run every channel at a fixed 1 kHz. You request a report rate R, and it picks the nearest supported ODR within

\n
.·R ≤ ConfiguredRate ≤ .·R\n
\n

up to each feature’s ceiling (e.g. raw accel 500 Hz, raw gyro 400 Hz, gyro-rotation vector 1 000 Hz, etc.). You enable whichever raw or fused streams you need, and the sensor hub programs its MEMS ODRs and MotionEngine fusion core accordingly.

\n
\n

First Comment

\n

Hi,
\nFrom https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/overview:

\n
\n

1x Inertial Measurement Unit (IMU; BNO080/BNO085): An integrated sensor that combines accelerometer, gyroscope, and magnetometer measurements with onboard processing to provide orientation and motion data.

\n
\n

The Helios IMU (BNO080/BNO085 SiP) seems to deliver a fixed 1 kHz sampling rate. In the product brief you’ll find:

\n
\n

“1 KHz Sample Rate – Enables flawless, smooth head-tracking with low latency…”

\n
\n

No alternate rates are listed, so you might consider the sample rate to be fixed at 1 kHz.

\n
\n

Update 1

\n

The competition linked pdf calls out the Helios IMU (BNO080/BNO085) as having a “1 kHz Sample Rate – Enables flawless, smooth head-tracking with low latency…” (from the product brief), which might suggest a fixed data rate.
\nHowever, the official BNO08X datasheet shows that all sensor and fusion outputs are configurable by the requested report rate R, and that the device picks the nearest supported Output Data Rate (ODR) such that

\n
. · R ≤ ConfiguredRate ≤ . · R\n
\n

Each feature has its own maximum ceiling:

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
Feature / SensorMax. Report RateFused?
Raw accelerometer500 HzNo
Raw gyroscope400 HzNo
Raw magnetometer100 HzNo
Gyro-rotation vector1 000 HzYes (gyro+accel+mag)
Rotation vector400 HzYes (accel+gyro+mag)
Gaming rotation vector400 HzYes (accel+gyro)
Geomagnetic rotation vector90 HzYes (accel+mag)
Gravity400 HzYes (accel+gyro)
Linear acceleration400 HzYes (accel+gyro)
\n

So rather than a hard-wired 1 kHz on all channels, you request whatever rate you need (up to each feature’s maximum) and the BNO08X will program the underlying MEMS ODRs and/or MotionEngine fusion core accordingly.

\n

—

\n

*Source: BNO080/085 Datasheet: https://www.ceva-ip.com/wp-content/uploads/BNO080_085-Datasheet.pdf*

", "votes": null, "replies": [ { "id": 3219930, "author_name": "defdet", "author_url": "", "post_date": "06/08/2025 14:26:55", "content": "

This would imply we have to classify the position based on observations that last ~0.1 second each

", "votes": null, "replies": [ { "id": 3219972, "author_name": "roysegalz", "author_url": "", "post_date": "06/08/2025 15:27:42", "content": "

@defdet Exacly what I thought… just doesn't add up.
\nLooks 50Hz'ish by the provided videos (about 1.5 second movements).
\nHope competition organizers will shed some light.

", "votes": null, "replies": [ { "id": 3220013, "author_name": "defdet", "author_url": "", "post_date": "06/08/2025 16:31:19", "content": "

yeah, it would be nice to receive some concrete answer, instead of having to speculate

", "votes": null, "replies": [ { "id": 3220028, "author_name": "maxstickel", "author_url": "", "post_date": "06/08/2025 17:06:21", "content": "

Well absolutely, I did not know how this IMU works ;) Sorry for bothering you.

", "votes": null, "replies": [] } ] } ] }, { "id": 3219979, "author_name": "maxstickel", "author_url": "", "post_date": "06/08/2025 15:41:24", "content": "

@roysegalz @defdet
\nUpdated my comment.
\nhttps://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583676#3219906

", "votes": null, "replies": [] } ] } ] }, { "id": 3220226, "author_name": "jaejohn", "author_url": "", "post_date": "06/09/2025 03:18:25", "content": "

as per the visualizer script, shared here: https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583118

\n
 ():\n    ...\n\n ():\n    ...\n    parser.add_argument(\n        , =, default=, =\n    )\n    ...\n
\n

the data can be interpreted at a sampling rate of 10 Hz (10 rows per second)

", "votes": null, "replies": [ { "id": 3227525, "author_name": "ogurtsov", "author_url": "", "post_date": "06/19/2025 04:26:58", "content": "

It looks reasonable, also it's completely in line with demo videos from competition page.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "3219904": "Hi, \n\nWhat is the sample rate on the IMU device?", "3219906": "**TL;DR:** Given the column names, we’re almost certainly looking at fused outputs. The BNO080/BNO085 doesn’t run every channel at a fixed 1 kHz. You request a report rate *R*, and it picks the nearest supported ODR within\n\n```\n0.9·R ≤ ConfiguredRate ≤ 2.1·R\n```\n\nup to each feature’s ceiling (e.g. raw accel 500 Hz, raw gyro 400 Hz, gyro-rotation vector 1 000 Hz, etc.). You enable whichever raw or fused streams you need, and the sensor hub programs its MEMS ODRs and MotionEngine fusion core accordingly.\n\n---\n\n## First Comment\nHi,\nFrom https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/overview:\n>1x Inertial Measurement Unit (IMU; BNO080/BNO085): An integrated sensor that combines accelerometer, gyroscope, and magnetometer measurements with onboard processing to provide orientation and motion data.\n\nThe Helios IMU (BNO080/BNO085 SiP) seems to deliver a fixed 1 kHz sampling rate. In the product brief you’ll find:\n> “1 KHz Sample Rate – Enables flawless, smooth head-tracking with low latency…”\n\nNo alternate rates are listed, so you might consider the sample rate to be fixed at 1 kHz.\n\n---\n## Update 1\nThe competition linked pdf calls out the Helios IMU (BNO080/BNO085) as having a “1 kHz Sample Rate – Enables flawless, smooth head-tracking with low latency…” (from the product brief), which might suggest a fixed data rate.\nHowever, the official BNO08X datasheet shows that **all** sensor and fusion outputs are *configurable* by the requested report rate R, and that the device picks the nearest supported Output Data Rate (ODR) such that\n\n```\n0.9 · R ≤ ConfiguredRate ≤ 2.1 · R\n```\n\nEach feature has its own maximum ceiling:\n\n| Feature / Sensor | Max. Report Rate | Fused? |\n| ------------------------------- | :--------------: | :------------------: |\n| **Raw accelerometer** | 500 Hz | No |\n| **Raw gyroscope** | 400 Hz | No |\n| **Raw magnetometer** | 100 Hz | No |\n| **Gyro-rotation vector** | 1 000 Hz | Yes (gyro+accel+mag) |\n| **Rotation vector** | 400 Hz | Yes (accel+gyro+mag) |\n| **Gaming rotation vector** | 400 Hz | Yes (accel+gyro) |\n| **Geomagnetic rotation vector** | 90 Hz | Yes (accel+mag) |\n| **Gravity** | 400 Hz | Yes (accel+gyro) |\n| **Linear acceleration** | 400 Hz | Yes (accel+gyro) |\n\nSo rather than a hard-wired 1 kHz on *all* channels, you request whatever rate you need (up to each feature’s maximum) and the BNO08X will program the underlying MEMS ODRs and/or MotionEngine fusion core accordingly.\n\n—\n\n*Source: BNO080/085 Datasheet: https://www.ceva-ip.com/wp-content/uploads/BNO080_085-Datasheet.pdf*", "3219930": "This would imply we have to classify the position based on observations that last ~0.1 second each", "3219972": "defdet Exacly what I thought... just doesn't add up.\nLooks 50Hz'ish by the provided videos (about 1.5 second movements).\nHope competition organizers will shed some light.", "3219979": "roysegalz @defdet \nUpdated my comment. \nhttps://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583676#3219906", "3220013": "yeah, it would be nice to receive some concrete answer, instead of having to speculate", "3220028": "Well absolutely, I did not know how this IMU works ;) Sorry for bothering you.", "3220226": "as per the visualizer script, shared here: https://www.kaggle.com/competitions/cmi-detect-behavior-with-sensor-data/discussion/583118\n\n```python\n\ndef render_animation(sequence_data, output_file, framerate=10):\n ...\n\ndef main():\n ...\n parser.add_argument(\n \"--framerate\", type=int, default=10, help=\"Frame rate for output video\"\n )\n ...\n```\n\nthe data can be interpreted at a sampling rate of 10 Hz (10 rows per second)", "3227525": "It looks reasonable, also it's completely in line with demo videos from competition page." }, "source": "meta" }