Images

def rerun.log_image(entity_path, image, *, draw_order=None, ext=None, timeless=False, recording=None, jpeg_quality=None)

Log a gray or color image.

The image should either have 1, 3 or 4 channels (gray, RGB or RGBA).

Supported dtypes

• uint8, uint16, uint32, uint64: color components should be in 0-max_uint sRGB gamma space, except for alpha which should be in 0-max_uint linear space.
• float16, float32, float64: all color components should be in 0-1 linear space.
• int8, int16, int32, int64: if all pixels are positive, they are interpreted as their unsigned counterparts. Otherwise, the image is normalized before display (the pixel with the lowest value is black and the pixel with the highest value is white).

Parameters:

Name	Type	Description	Default
entity_path	str	Path to the image in the space hierarchy.	required
image	Tensor	A Tensor representing the image to log.	required
draw_order	float | None	An optional floating point value that specifies the 2D drawing order. Objects with higher values are drawn on top of those with lower values. The default for images is -10.0.	None
ext	dict[str, Any] | None	Optional dictionary of extension components. See rerun.log_extension_components	None
timeless	bool	If true, the image will be timeless (default: False).	False
recording	RecordingStream | None	Specifies the rerun.RecordingStream to use. If left unspecified, defaults to the current active data recording, if there is one. See also: rerun.init, rerun.set_global_data_recording.	None
jpeg_quality	int | None	If set, encode the image as a JPEG to save storage space. Higher quality = larger file size. A quality of 95 still saves a lot of space, but is visually very similar. JPEG compression works best for photographs. Only RGB images are supported. Note that compressing to JPEG costs a bit of CPU time, both when logging and later when viewing them.	None

def rerun.log_image_file(entity_path, *, img_bytes=None, img_path=None, img_format=None, timeless=False, recording=None)

Log an image file given its contents or path on disk.

You must pass either img_bytes or img_path.

Only JPEGs and PNGs are supported right now.

JPEGs will be stored compressed, saving memory, whilst PNGs will currently be decoded before they are logged. This may change in the future.

If no img_format is specified, rerun will try to guess it.

Parameters:

Name	Type	Description	Default
entity_path	str	Path to the image in the space hierarchy.	required
img_bytes	bytes | None	Content of an image file, e.g. a .jpg.	None
img_path	Path | None	Path to an image file, e.g. a .jpg.	None
img_format	ImageFormat | None	Format of the image file.	None
timeless	bool	If true, the image will be timeless (default: False).	False
recording	RecordingStream | None	Specifies the rerun.RecordingStream to use. If left unspecified, defaults to the current active data recording, if there is one. See also: rerun.init, rerun.set_global_data_recording.	None

def rerun.log_depth_image(entity_path, image, *, draw_order=None, meter=None, ext=None, timeless=False, recording=None)

Log a depth image.

The image must be a 2D array.

Supported dtypes

float16, float32, float64, uint8, uint16, uint32, uint64, int8, int16, int32, int64

Parameters:

Name	Type	Description	Default
entity_path	str	Path to the image in the space hierarchy.	required
image	Tensor	A Tensor representing the depth image to log.	required
draw_order	float | None	An optional floating point value that specifies the 2D drawing order. Objects with higher values are drawn on top of those with lower values. The default for images is -10.0.	None
meter	float | None	How long is a meter in the given dtype? For instance: with uint16, perhaps meter=1000 which would mean you have millimeter precision and a range of up to ~65 meters (2^16 / 1000).	None
ext	dict[str, Any] | None	Optional dictionary of extension components. See rerun.log_extension_components	None
timeless	bool	If true, the image will be timeless (default: False).	False
recording	RecordingStream | None	Specifies the rerun.RecordingStream to use. If left unspecified, defaults to the current active data recording, if there is one. See also: rerun.init, rerun.set_global_data_recording.	None

def rerun.log_segmentation_image(entity_path, image, *, draw_order=None, ext=None, timeless=False, recording=None)

Log an image made up of integer class-ids.

The image should have 1 channel, i.e. be either H x W or H x W x 1.

See: rerun.log_annotation_context for information on how to map the class-ids to colors and labels.

Supported dtypes

uint8, uint16

Parameters:

Name	Type	Description	Default
entity_path	str	Path to the image in the space hierarchy.	required
image	ArrayLike	A Tensor representing the segmentation image to log.	required
draw_order	float | None	An optional floating point value that specifies the 2D drawing order. Objects with higher values are drawn on top of those with lower values. The default for images is -10.0.	None
ext	dict[str, Any] | None	Optional dictionary of extension components. See rerun.log_extension_components	None
timeless	bool	If true, the image will be timeless (default: False).	False
recording	RecordingStream | None	Specifies the rerun.RecordingStream to use. If left unspecified, defaults to the current active data recording, if there is one. See also: rerun.init, rerun.set_global_data_recording.	None