Video input requirements
In order to perform a qualitative assessment of the health parameters from a video, it is necessary to adhere to the following input requirements for the video recorded from the subject.
As a commitment to maintaining a quality-first approach, we refrain from delivering results when accuracy cannot be guaranteed due to unmet input requirements. Our engine will return a clear error_code in this case, explaining which input requirement was not met. A list of possible error_codes can be found here. In these specific cases we advise to not abrubtly stop the end-user flow, but gracefully handle the error and inform the user on how to perform the measurement correctly.
It is important for the subject to remain still during the recording and center the face within the recording frame. The goal is to keep looking at the camera and minimize head movement. The setup is similar to when taking a photo.
For comprehensive guidelines on how to correctly sit in front of the camera during the measurement, we refer to our Usage instructions.
It is important to perform the recording in a well lit room. A lighting of at least 200 lux is recommended.
The recording length of the video should be at least 20 seconds and should not exceed a length of one minute. A video length of 30 seconds is recommended.
A video length of 30 seconds provides a nice trade-off between accuracy, processing time and user experience.
Video's with a duration less then 20 seconds will not be processed. More info can be found here: API docs > Get results > video too short.
The video frame rate should at least be 30 fps and should not exceed 60 fps. The recommended value is a recording frame rate of 30 fps.
Higher frame rates may lead to a marginal increase in accuracy, but highly increases upload and processing time.
The video resolution, codec, and bitrate are critical properties to consider when capturing and processing video data for measuring the biomarkers using our algorithm. These factors play a vital role in preserving the necessary information about the blood volume pulse, which is essential for accurate measurements.
- Resolution: between 320p and 1080p.
- Codec: H264 or H265 if supported; VP8 when recording in most browsers.
- Bitrate: depending on resolution and codec, see table below.
Below you find a reference table with the minimal bitrate values (expressed in kbps) to which videos must comply. Mind that the minimal bitrate is both depending on the resolution and video codec.
Bitrate (kbps) | H264 | H265 |
|---|---|---|
320p | 4700 | 4030 |
480p | 1030 | 1280 |
640p | 2320 | 1540 |
720p | 1220 | 1940 |
1080p | 850 | 2890 |
In case you are not sure the video is recorded correctly, we recommend using a tool like https://www.metadata2go.com to verify the video's metadata. The r_frame_rate field should contain a value between 30 and 60 fps.
The camera sensor is the part of the imaging system that is sensitive to light. The size of this sensor plays an important role in the final quality of the image or video. The larger the sensor, the more sensitive it is to light. Larger image sensors have larger pixels, which means less noise, better dynamic range and the ability to obtain more (subtle) information. The size of the sensor should be at least 5.0 mm in diagonal. The size of a pixel (unit cell) should be at least 1.0 μm.
Lens quality is becoming more important with the increasing pixel count on modern cameras. The resolution of digital photos is often limited by the camera lens rather than the resolution of the camera itself. In order to perform a qualitative measurement, the lens should have an aperture of at least f2.8 and be made from glass or an (optically) higher quality material.