Depth Compute
The Depth Compute CLI processes data from the Data Collect CLI to generate three types of output frames.
Important
This tool has been removed from the host starting with Eval Kit version 5.0.0.
Output Folders
RadialDepth
Depth images where the depth at each pixel is the distance to the corresponding point in the scene measured ‘radially’ from the center of the lens.
AB (Active Brightness)
Active Brightness images showing laser return intensity from scene points (IR frame).
XYZ
Point cloud data with real-world Cartesian coordinates (X, Y, Z) for each pixel. The Z image alone functions as a distance measurement from the camera plane.
Usage
EVAL-ADTF3175-NXZ
tofi_compute_depth.exe --I=../data_output --CCB=../crXXX.ccb --MODE=10 --O=../proc_data --INI=../config/RawToDepth.ini
Parameters:
--I: Input data folder--CCB: Camera calibration file--MODE: Mode index (10 for mp, 7 for qmp)--O: Output folder--INI: INI configuration file
Visualization
Use the provided Python scripts to visualize the output:
Visualize Depth
python tools\depth_compute\visualize_depth.py proc_data\RadialDepth\data_output_0.bin 1024 1024
Visualize Active Brightness
python tools\depth_compute\visualize_ab.py proc_data\AB\data_output_0.bin 1024 1024
Visualize Point Cloud
python tools\depth_compute\visualize_pointcloud.py 1024 1024 proc_data\XYZ\data_output_0.bin
Parameters
Users adjust depth compute parameters through an initialization INI file during SDK runtime. This file also controls ADSD3500 settings when applicable.
Recommended Settings
Example baseline configuration:
[depth-compute]
abThreshMin = 3.0
confThresh = 25.0
radialThreshMin = 100
radialThreshMax = 10000
jblfApplyFlag = 1
jblfWindowSize = 7
jblfGaussianSigma = 10
Parameter Reference
Parameter |
Unit |
Description |
|---|---|---|
abThreshMin |
– |
Per-frequency brightness invalidation threshold. Ranges across 16-bit AB bitwidth. |
confThresh |
degrees |
Phase unwrapping confidence threshold (0-360 degrees). |
radialThreshMin |
mm |
Minimum distance-based invalidation boundary. |
radialThreshMax |
mm |
Maximum distance-based invalidation boundary. |
jblfApplyFlag |
bool |
Enables (1) or disables (0) noise reduction filtering. |
jblfWindowSize |
pixels |
Filter kernel dimension. Options: 3, 5, 7. |
jblfGaussianSigma |
– |
Gaussian sigma value for the joint bilateral filter. |
jblfABThreshold |
– |
Active brightness threshold for the filter. |
jblfExponentialTerm |
– |
Exponential term for the filter calculation. |
jblfMaxEdge |
– |
Maximum edge value for filtering. |
partialDepthEnable |
bool |
Enable partial depth computation. |
phaseInvalid |
– |
Phase invalid marker value. |
bitsInPhaseOrDepth |
bits |
Number of bits allocated for phase or depth data. |
ADSD3500 Configuration Parameters
Parameter |
Description |
|---|---|
inputFormat |
Input data format specification |
bitsInAB |
Bits allocated for active brightness |
bitsInConf |
Bits allocated for confidence data |
interleavingEnable |
Enable interleaving mode |
xyzEnable |
Enable XYZ frame generation |
Input Format Options
Eight input format variants are available:
Format |
Description |
|---|---|
mipiRaw12 |
Standard MIPI RAW12 format |
mipiRaw12_8 |
MIPI RAW12 with 8-bit packing |
raw8 |
Compressed RAW8 format |
raw16 |
Standard RAW16 format |
raw16_unpacked |
Unpacked RAW16 for NXP interface |
raw16_unpacked_high |
High bits unpacked RAW16 |
raw16_unpacked_low |
Low bits unpacked RAW16 |
compressed_raw8 |
Compressed 8-bit format |
Libraries
The ADTF3175D kits require depth compute binaries to convert raw data to depth data. These binaries are necessary for the SDK to function properly during the build process.
The installers are located within an image download package available through the GitHub installer. See Software Installation for download instructions.
Windows Binaries
Version |
Type |
Description |
|---|---|---|
X.X.0 |
GPU-accelerated |
OpenCL version using GPU acceleration |
X.X.1 |
CPU-based |
OpenCL version using CPU (default in GitHub Installer 3.2.0) |
Linux Binaries
Version |
Type |
Description |
|---|---|---|
X.X.0 |
GPU-accelerated |
OpenCL version using GPU acceleration |
X.X.1 |
CPU-based |
OpenCL version using CPU |
Selection Guide
Use GPU-accelerated version (X.X.0) if your system has a compatible GPU with OpenCL support for best performance.
Use CPU-based version (X.X.1) if you don’t have a compatible GPU or experience issues with the GPU version.
Note
The CPU-based version is set as the default in GitHub Installer 3.2.0 and later for maximum compatibility.