Skip to content

Holofile Source Task

This task reads data from a .holo file and outputs it as a tensor. When the end of the file is reached, it loops back to the beginning. The task can be configured to load data in different ways, such as loading all frames into memory at once or loading frames on-the-fly. The file specified must exist and be a valid .holo file during the task's validation phase and execution.

Inputs

This task does not have any inputs.

Outputs

This task has a single output of shape (B, H, W), where:

  • B is the batch size specified in the settings.
  • H and W are the height and width of the frames in the .holo file.

The dtype of the output tensor is unsigned N-bit integer, where N is the bits per pixel specified in the file header. If the file bits per pixel is not 8, 16, or 32, the output tensor will have dtype uint8.

The memory location of the output tensor is determined by the load_kind setting. If load_kind is set to Live or CPUCached, the output tensor will be in host memory. If load_kind is set to GPUCached, the output tensor will be in device memory.

Inplace

This task does not support inplace operation.

Ownership

This task does not own any inputs or outputs.

Settings

Example:

1
2
3
4
5
6
7
{
    "path": "C:/data/run01.holo",
    "load_kind": "Live",
    "start_frame": 0,
    "end_frame": 1024,
    "batch_size": 8
}
Property Pattern Type Deprecated Definition Title/Description
+ path No string No - Path to the HoloFile. Must be a non-empty string.
+ load_kind No enum (of string) No - Loading strategy. Accepted values:
- Live: Read on demand from disk.
- CPUCached: Preload all frames into CPU memory.
- GPUCached: Preload all frames into GPU memory.
+ start_frame No integer No - First frame to read (inclusive). Must satisfy:
- start_frame ≥ 0
- start_frame < number of frames in file
+ end_frame No integer No - Last frame to read (exclusive). Must satisfy:
- end_frame ≥ start_frame
- end_frame - start_frame ≥ batch_size
- end_frame <= number of frames in file
+ batch_size No integer No - Number of frames per output tensor. Must satisfy:
- batch_size > 0
- batch_size ≤ (end_frame - start_frame)