Dear Luxonis Group,
I am trying to send out a depthmap with my ToF sensor through spout in 16 oder 32 bit
and would like to convert my Depthmap stream in Touchdesigner into a Pointcloud.
Somehow changing the unit8 to 32 or something similar kills my code.
Can anybody have a look and maybe tell me at white codelines I could do the change to send out a 32bit spout stream?
Thank you very much for your help 
BonkoKaradjov Apparently spout typically expects RGBA 8bit floats so pushing 32bit values through will not work.
If you are using some specific framework that is capable of 32bit - FROM GPT:
To output a 16-bit or 32-bit depth map (instead of an 8-bit colorized image), you need to remove or bypass the colorizeDepth step and ensure that your Syphon texture is created/handled with a high-bit-depth format. Below is a conceptual outline:
Remove the colorizeDepth call:
# alignedDepth = colorizeDepth(alignedDepth)Convert your raw depth data (in millimeters) to float:
alignedDepth_32 = alignedDepth.astype(np.float32)Now you have a single-channel float array with shape (height, width).
Expand to 4 channels (RGBA) if your Syphon/Metal functions require it:
h, w = alignedDepth_32.shape
alignedDepth_32_rgba = np.zeros((h, w, 4), dtype=np.float32)
alignedDepth_32_rgba[..., 0] = alignedDepth_32 # R
alignedDepth_32_rgba[..., 1] = alignedDepth_32 # G
alignedDepth_32_rgba[..., 2] = alignedDepth_32 # B
alignedDepth_32_rgba[..., 3] = 1.0 # ACreate the Metal texture in a float format if possible. For example:
texture_tof = create_mtl_texture(
syphon_server_tof.device,
640,
480,
pixel_format="RGBA32Float" # or a similar float-friendly format
)(Exact format strings may vary based on your Syphon library.)
Copy the float data to that texture:
copy_image_to_mtl_texture(alignedDepth_32_rgba, texture_tof)Publish via Syphon:
syphon_server_tof.publish_frame_texture(texture_tof)This way, you’re sending a raw float texture rather than an 8-bit color map, which allows TouchDesigner (or another application) to interpret the depth data for building a point cloud.
Thanks,
Jaka
Dear @jakaskerl ,
thank you very much for your help.
Mostly I also use gpt to look up some answers, sometimes the code is confused because of libraries gpt makes of 
In this case, I could achieve somehow a inbetween send but this is my outcome
and decided to go through all possible metal colorspaces:
Sadly my stream not seems to be 32bit and gives out some interesting videostreams, I
don't know which one could be the right one to create a pointlcloud into TD or if I am doing something else wrong.
Link to a video preview switching all possible color spaces from mtlpixeformat website
(( Website with mptlpixelformats ))
Thank again you very much for your help
BonkoKaradjov
Was this device calibrated? In the video, it looks like you are manually setting the offset.. Could you sent the calibration dump so we can check?
Also the code you sent looks different from the one used to create the video.
Thanks,
Jaka
Dear @jakaskerl ,
I had sent my 1MP ToF sensor to Luxonis and payed extra 300€ with the goal to get a aligned 12MP camera to ToF sensor and send out, as in my last post a (minimum) 1352x1014 for RGB + minimum 640x480 (ToF) stream via ndi or syphon.
If I got it right Luxonis answered me that the sensor should had been calibrated.
Therefore I am scared, because of my already running production to do something wrong if I run the calibration script from Luxonis, that is on the website.
My hope was to get a as perfect as possible alignement (when person gets near, left, right, and far from the cam etc.) between both cameras with minumum latency via PoE.
Everything, except the alignement (and the point cloud data for Touchdesigner) out from the I could get with these two scripts that I worked out for a long time, because I am slow in coding:
The alignment problem I really tryed in so many scripts and changes and also had emails with Erik who really tried to help me.
I have maybe over 30-40 tries and scripts that I modified.
The following links show videos where I bound in the original scripts you asked me to implement without the custom transform:
ToF-Alignment Script Video from Luxonis Website (original)
Link to Script
ToF-Alignment Script Video from Luxonis Website (rotated 180 degrees after alignment)
Link to Script
ToF-Alignment Script sent from Luxonis by Mail (original)
Link to Script
- one unknown error: freezes often after a few seconds for 30-60 seconds
ToF-Alignment Script sent from Luxonis by Mail (rotated 180 degrees after alignment)
Link to Script
- one unknown error: freezes often after a few seconds for 30-60 seconds
I will try 3 or 4 more days if I can find at least my workaround in aligning the cameras in Touchdesigner and somehow sending out the ndi/syphon pointcloud stream data to create a pointcloud in TD and if I can not use the sensor for the production I hope Luxonis can give me a refund for the sensor and the custom built so that I can switch to my first attempt with the Orbecc Femto Mega because my skills seam to be to bad to work with Luxonis sensors at the moment, although it was quite fun to start to understand them and have small successes but it feels wrong to pay 750€ for the Oak ToF 12MP and work almost 30 days on trying to get the alignment right etc.
Thank you very much for your help and sorry that I am so complicated, its just my lack of knowledge in coding with depthai and opencv etc. and that I am disappointed of myself that I will probably not achieve the simple goal I could get with Kinect or Femto Mega in a much easier way and that I invested so much money, almost the whole money I will get for this art project I need to use this sensor for.
@jakaskerl And here the calibration dump
....main/examples/calibration/calibration_dump.py
Is EEPROM available: True
User calibration: {
"batchName": "",
"batchTime": 1724403929,
"boardConf": "IR-C80M00-00",
"boardCustom": "",
"boardName": "EL2086",
"boardOptions": 6401,
"boardRev": "R3M2E3",
"cameraData": [
[
2,
{
"cameraType": 0,
"distortionCoeff": [
-1.906318187713623,
2.000298500061035,
-0.002747641410678625,
-0.0008416209020651877,
-0.512755811214447,
-2.004528284072876,
2.276949644088745,
-0.7380069494247437,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.9999335408210754,
-0.008848309516906738,
0.007388767320662737
],
[
0.008809356018900871,
0.999947190284729,
0.00528801279142499
],
[
-0.007435167208313942,
-0.0052225710824131966,
0.9999586939811707
]
],
"specTranslation": {
"x": 1.7381999492645264,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": 0,
"translation": {
"x": 1.9038344621658325,
"y": 0.16774022579193115,
"z": 0.18684668838977814
}
},
"height": 3040,
"intrinsicMatrix": [
[
3114.211669921875,
0.0,
2067.387939453125
],
[
0.0,
3120.09423828125,
1523.5755615234375
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 4056
}
],
[
1,
{
"cameraType": 0,
"distortionCoeff": [
22.9776611328125,
-35.2191276550293,
-0.0008518987451680005,
-0.0005938760004937649,
13.29187297821045,
22.328569412231445,
-33.41155242919922,
11.978973388671875,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.999860405921936,
0.007616722956299782,
-0.014872560277581215
],
[
-0.007512797601521015,
0.9999470710754395,
0.007031152956187725
],
[
0.014925327152013779,
-0.006918436847627163,
0.9998646974563599
]
],
"specTranslation": {
"x": 2.0,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": 2,
"translation": {
"x": 1.988970160484314,
"y": -0.0013319612480700016,
"z": 0.0029759984463453293
}
},
"height": 800,
"intrinsicMatrix": [
[
801.0270385742188,
0.0,
609.644287109375
],
[
0.0,
802.9794921875,
392.1715393066406
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 1280
}
],
[
0,
{
"cameraType": 0,
"distortionCoeff": [
-1.9164514541625977,
-29.196409225463867,
-0.0034538470208644867,
-0.0005991361103951931,
87.63078308105469,
-1.9397814273834229,
-29.01816177368164,
87.28975677490234,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
]
],
"specTranslation": {
"x": -0.0,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": -1,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"height": 480,
"intrinsicMatrix": [
[
473.2501525878906,
0.0,
319.0169982910156
],
[
0.0,
473.8380432128906,
228.83583068847656
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 640
}
]
],
"deviceName": "",
"hardwareConf": "F1-FV00-BC000",
"housingExtrinsics": {
"rotationMatrix": [
[
0.9998946189880371,
0.0140096265822649,
0.00380001706071198
],
[
-0.014019245281815529,
0.9998985528945923,
0.0025167076382786036
],
[
-0.0037643732503056526,
-0.0025697157252579927,
0.9999896287918091
]
],
"specTranslation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
},
"toCameraSocket": 0,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"imuExtrinsics": {
"rotationMatrix": [
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
]
],
"specTranslation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
},
"toCameraSocket": -1,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"miscellaneousData": [],
"productName": "FFC-4P",
"stereoEnableDistortionCorrection": true,
"stereoRectificationData": {
"leftCameraSocket": 1,
"rectifiedRotationLeft": [
[
0.9998863935470581,
0.006936723832041025,
-0.013381202705204487
],
[
-0.006890007760375738,
0.999970018863678,
0.003534123534336686
],
[
0.01340531650930643,
-0.0034415253903716803,
0.999904215335846
]
],
"rectifiedRotationRight": [
[
0.9999986290931702,
-0.0006696729105897248,
0.0014962489949539304
],
[
0.0006748876767233014,
0.9999936819076538,
-0.0034874198026955128
],
[
-0.0014939040411263704,
0.0034884249325841665,
0.9999927878379822
]
],
"rightCameraSocket": 2
},
"stereoUseSpecTranslation": false,
"version": 7,
"verticalCameraSocket": -1
}
Factory calibration: {
"batchName": "",
"batchTime": 1724403929,
"boardConf": "IR-C80M00-00",
"boardCustom": "",
"boardName": "EL2086",
"boardOptions": 6401,
"boardRev": "R3M2E3",
"cameraData": [
[
2,
{
"cameraType": 0,
"distortionCoeff": [
-1.906318187713623,
2.000298500061035,
-0.002747641410678625,
-0.0008416209020651877,
-0.512755811214447,
-2.004528284072876,
2.276949644088745,
-0.7380069494247437,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.9999335408210754,
-0.008848309516906738,
0.007388767320662737
],
[
0.008809356018900871,
0.999947190284729,
0.00528801279142499
],
[
-0.007435167208313942,
-0.0052225710824131966,
0.9999586939811707
]
],
"specTranslation": {
"x": 1.7381999492645264,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": 0,
"translation": {
"x": 1.9038344621658325,
"y": 0.16774022579193115,
"z": 0.18684668838977814
}
},
"height": 3040,
"intrinsicMatrix": [
[
3114.211669921875,
0.0,
2067.387939453125
],
[
0.0,
3120.09423828125,
1523.5755615234375
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 4056
}
],
[
1,
{
"cameraType": 0,
"distortionCoeff": [
22.9776611328125,
-35.2191276550293,
-0.0008518987451680005,
-0.0005938760004937649,
13.29187297821045,
22.328569412231445,
-33.41155242919922,
11.978973388671875,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.999860405921936,
0.007616722956299782,
-0.014872560277581215
],
[
-0.007512797601521015,
0.9999470710754395,
0.007031152956187725
],
[
0.014925327152013779,
-0.006918436847627163,
0.9998646974563599
]
],
"specTranslation": {
"x": 2.0,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": 2,
"translation": {
"x": 1.988970160484314,
"y": -0.0013319612480700016,
"z": 0.0029759984463453293
}
},
"height": 800,
"intrinsicMatrix": [
[
801.0270385742188,
0.0,
609.644287109375
],
[
0.0,
802.9794921875,
392.1715393066406
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 1280
}
],
[
0,
{
"cameraType": 0,
"distortionCoeff": [
-1.9164514541625977,
-29.196409225463867,
-0.0034538470208644867,
-0.0005991361103951931,
87.63078308105469,
-1.9397814273834229,
-29.01816177368164,
87.28975677490234,
0.0,
0.0,
0.0,
0.0,
0.0,
0.0
],
"extrinsics": {
"rotationMatrix": [
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
]
],
"specTranslation": {
"x": -0.0,
"y": -0.0,
"z": -0.0
},
"toCameraSocket": -1,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"height": 480,
"intrinsicMatrix": [
[
473.2501525878906,
0.0,
319.0169982910156
],
[
0.0,
473.8380432128906,
228.83583068847656
],
[
0.0,
0.0,
1.0
]
],
"lensPosition": 0,
"specHfovDeg": 71.86000061035156,
"width": 640
}
]
],
"deviceName": "",
"hardwareConf": "F1-FV00-BC000",
"housingExtrinsics": {
"rotationMatrix": [
[
0.9998946189880371,
0.0140096265822649,
0.00380001706071198
],
[
-0.014019245281815529,
0.9998985528945923,
0.0025167076382786036
],
[
-0.0037643732503056526,
-0.0025697157252579927,
0.9999896287918091
]
],
"specTranslation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
},
"toCameraSocket": 0,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"imuExtrinsics": {
"rotationMatrix": [
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
],
[
0.0,
0.0,
0.0
]
],
"specTranslation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
},
"toCameraSocket": -1,
"translation": {
"x": 0.0,
"y": 0.0,
"z": 0.0
}
},
"miscellaneousData": [],
"productName": "FFC-4P",
"stereoEnableDistortionCorrection": true,
"stereoRectificationData": {
"leftCameraSocket": 1,
"rectifiedRotationLeft": [
[
0.9998863935470581,
0.006936723832041025,
-0.013381202705204487
],
[
-0.006890007760375738,
0.999970018863678,
0.003534123534336686
],
[
0.01340531650930643,
-0.0034415253903716803,
0.999904215335846
]
],
"rectifiedRotationRight": [
[
0.9999986290931702,
-0.0006696729105897248,
0.0014962489949539304
],
[
0.0006748876767233014,
0.9999936819076538,
-0.0034874198026955128
],
[
-0.0014939040411263704,
0.0034884249325841665,
0.9999927878379822
]
],
"rightCameraSocket": 2
},
"stereoUseSpecTranslation": false,
"version": 7,
"verticalCameraSocket": -1
}
User calibration raw: [7, 0, 170, 85, 3, 0, 0, 0, 0, 69, 76, 50, 48, 56, 54, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 51, 77, 50, 69, 51, 0, 0, 0, 0, 142, 248, 127, 63, 117, 77, 227, 59, 213, 60, 91, 188, 147, 197, 225, 187, 9, 254, 127, 63, 193, 156, 103, 59, 249, 161, 91, 60, 55, 139, 97, 187, 185, 249, 127, 63, 233, 255, 127, 63, 253, 140, 47, 186, 201, 29, 196, 58, 242, 234, 48, 58, 150, 255, 127, 63, 50, 141, 100, 187, 26, 207, 195, 186, 15, 158, 100, 59, 135, 255, 127, 63, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 32, 3, 0, 5, 0, 187, 65, 72, 68, 0, 0, 0, 0, 60, 105, 24, 68, 0, 0, 0, 0, 176, 190, 72, 68, 245, 21, 196, 67, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 64, 210, 183, 65, 99, 224, 12, 194, 245, 81, 95, 186, 88, 174, 27, 186, 131, 171, 84, 65, 233, 160, 178, 65, 110, 165, 5, 194, 224, 169, 63, 65, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 1, 218, 246, 127, 63, 180, 149, 249, 59, 10, 172, 115, 188, 234, 45, 246, 187, 136, 252, 127, 63, 150, 101, 230, 59, 92, 137, 116, 60, 14, 180, 226, 187, 34, 247, 127, 63, 147, 150, 254, 63, 52, 149, 174, 186, 248, 8, 67, 59, 0, 0, 0, 192, 0, 0, 0, 0, 0, 0, 0, 0, 2, 224, 1, 128, 2, 0, 5, 160, 236, 67, 0, 0, 0, 0, 45, 130, 159, 67, 0, 0, 0, 0, 69, 235, 236, 67, 249, 213, 100, 67, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 72, 78, 245, 191, 63, 146, 233, 193, 240, 89, 98, 187, 88, 15, 29, 186, 246, 66, 175, 66, 194, 74, 248, 191, 50, 37, 232, 193, 91, 148, 174, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 224, 11, 216, 15, 0, 99, 163, 66, 69, 0, 0, 0, 0, 53, 54, 1, 69, 0, 0, 0, 0, 130, 1, 67, 69, 107, 114, 190, 68, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 60, 2, 244, 191, 228, 4, 0, 64, 198, 17, 52, 187, 57, 160, 92, 186, 247, 67, 3, 191, 49, 74, 0, 192, 139, 185, 17, 64, 6, 238, 60, 191, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 2, 165, 251, 127, 63, 128, 248, 16, 188, 121, 29, 242, 59, 30, 85, 16, 60, 138, 252, 127, 63, 17, 71, 173, 59, 180, 162, 243, 187, 26, 34, 171, 187, 75, 253, 127, 63, 217, 176, 243, 63, 24, 196, 43, 62, 189, 84, 63, 62, 86, 125, 222, 191, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 70, 70, 67, 45, 52, 80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 73, 82, 45, 67, 56, 48, 77, 48, 48, 45, 48, 48, 0, 0, 0, 0, 70, 49, 45, 70, 86, 48, 48, 45, 66, 67, 48, 48, 48, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 217, 80, 200, 102, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 25, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 24, 249, 127, 63, 162, 136, 101, 60, 181, 9, 121, 59, 250, 176, 101, 188, 90, 249, 127, 63, 89, 239, 36, 59, 180, 179, 118, 187, 173, 104, 40, 187, 82, 255, 127, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 3, 255]
Factory calibration raw: [7, 0, 170, 85, 3, 0, 0, 0, 0, 69, 76, 50, 48, 56, 54, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 51, 77, 50, 69, 51, 0, 0, 0, 0, 142, 248, 127, 63, 117, 77, 227, 59, 213, 60, 91, 188, 147, 197, 225, 187, 9, 254, 127, 63, 193, 156, 103, 59, 249, 161, 91, 60, 55, 139, 97, 187, 185, 249, 127, 63, 233, 255, 127, 63, 253, 140, 47, 186, 201, 29, 196, 58, 242, 234, 48, 58, 150, 255, 127, 63, 50, 141, 100, 187, 26, 207, 195, 186, 15, 158, 100, 59, 135, 255, 127, 63, 1, 2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 32, 3, 0, 5, 0, 187, 65, 72, 68, 0, 0, 0, 0, 60, 105, 24, 68, 0, 0, 0, 0, 176, 190, 72, 68, 245, 21, 196, 67, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 64, 210, 183, 65, 99, 224, 12, 194, 245, 81, 95, 186, 88, 174, 27, 186, 131, 171, 84, 65, 233, 160, 178, 65, 110, 165, 5, 194, 224, 169, 63, 65, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 1, 218, 246, 127, 63, 180, 149, 249, 59, 10, 172, 115, 188, 234, 45, 246, 187, 136, 252, 127, 63, 150, 101, 230, 59, 92, 137, 116, 60, 14, 180, 226, 187, 34, 247, 127, 63, 147, 150, 254, 63, 52, 149, 174, 186, 248, 8, 67, 59, 0, 0, 0, 192, 0, 0, 0, 0, 0, 0, 0, 0, 2, 224, 1, 128, 2, 0, 5, 160, 236, 67, 0, 0, 0, 0, 45, 130, 159, 67, 0, 0, 0, 0, 69, 235, 236, 67, 249, 213, 100, 67, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 72, 78, 245, 191, 63, 146, 233, 193, 240, 89, 98, 187, 88, 15, 29, 186, 246, 66, 175, 66, 194, 74, 248, 191, 50, 37, 232, 193, 91, 148, 174, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 224, 11, 216, 15, 0, 99, 163, 66, 69, 0, 0, 0, 0, 53, 54, 1, 69, 0, 0, 0, 0, 130, 1, 67, 69, 107, 114, 190, 68, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 128, 63, 60, 2, 244, 191, 228, 4, 0, 64, 198, 17, 52, 187, 57, 160, 92, 186, 247, 67, 3, 191, 49, 74, 0, 192, 139, 185, 17, 64, 6, 238, 60, 191, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 82, 184, 143, 66, 0, 2, 165, 251, 127, 63, 128, 248, 16, 188, 121, 29, 242, 59, 30, 85, 16, 60, 138, 252, 127, 63, 17, 71, 173, 59, 180, 162, 243, 187, 26, 34, 171, 187, 75, 253, 127, 63, 217, 176, 243, 63, 24, 196, 43, 62, 189, 84, 63, 62, 86, 125, 222, 191, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 70, 70, 67, 45, 52, 80, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 73, 82, 45, 67, 56, 48, 77, 48, 48, 45, 48, 48, 0, 0, 0, 0, 70, 49, 45, 70, 86, 48, 48, 45, 66, 67, 48, 48, 48, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 217, 80, 200, 102, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 25, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 24, 249, 127, 63, 162, 136, 101, 60, 181, 9, 121, 59, 250, 176, 101, 188, 90, 249, 127, 63, 89, 239, 36, 59, 180, 179, 118, 187, 173, 104, 40, 187, 82, 255, 127, 63, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 255, 3, 255]
Process finished with exit code 0BonkoKaradjov
Thanks, can you also send a left camera frame and ToF depth frame so we can test your calibration on our side?
edit: images should be taken at the same time of course
edit2: i see I'm mixing the threads a bit, sorry for that, let's first fix alignment and the move on to the depth streaming
Thanks,
Jaka
Dear @jakaskerl,
thank you very much for your help.
To get a better quality of the maps I could record the images in Touchdesigner through sending ndi for example of the third additional stream (cam_a, cam_b, and cam_c).
Or record it directly from hard disk, this would probably give you the most exact output.
But I am not shure how to implement the third camera synced to the other two to produce the output.
Is there any script example out there that I can use for this with the ToF sensor to record or send out the
unaligned 3 videostreams through syphon, ndi or record on disk?
I am slow in coding such things and it would be really helpful if there is a code example instead of implementing this into the two scripts above and probably sending you the wrong data.
I would also switch the space where I record the video so that you get more angles and data in the depth.
Thank you very much for your help.
With kind regards
Bonko
BonkoKaradjov
Should be good:
import os
import time
import cv2
import depthai as dai
import numpy as np
from datetime import timedelta
def create_pipeline():
pipeline = dai.Pipeline()
# Create ToF node
tof = pipeline.create(dai.node.ToF)
tof.setNumShaves(4)
# Configure ToF
tofConfig = tof.initialConfig.get()
tofConfig.enableFPPNCorrection = True
tofConfig.enableOpticalCorrection = True
tofConfig.enableWiggleCorrection = True
tofConfig.enableTemperatureCorrection = True
tofConfig.phaseUnwrappingLevel = 4
tof.initialConfig.set(tofConfig)
# Create ToF camera node
cam_tof = pipeline.create(dai.node.Camera)
cam_tof.setFps(20)
cam_tof.setImageOrientation(dai.CameraImageOrientation.ROTATE_180_DEG)
cam_tof.setBoardSocket(dai.CameraBoardSocket.CAM_A)
cam_tof.raw.link(tof.input)
# Create RGB camera nodes
colorLeft = pipeline.create(dai.node.ColorCamera)
colorRight = pipeline.create(dai.node.ColorCamera)
# Configure RGB cameras
colorLeft.setBoardSocket(dai.CameraBoardSocket.CAM_B)
colorRight.setBoardSocket(dai.CameraBoardSocket.CAM_C)
colorLeft.setResolution(dai.ColorCameraProperties.SensorResolution.THE_800_P)
colorRight.setResolution(dai.ColorCameraProperties.SensorResolution.THE_800_P)
colorLeft.setFps(20)
colorRight.setFps(20)
colorLeft.setInterleaved(False)
colorRight.setInterleaved(False)
# Create Sync node
sync = pipeline.create(dai.node.Sync)
sync.setSyncThreshold(timedelta(milliseconds=50))
# Link outputs to Sync
tof.depth.link(sync.inputs["tof"])
colorLeft.isp.link(sync.inputs["left"])
colorRight.isp.link(sync.inputs["right"])
# Create output
xout = pipeline.create(dai.node.XLinkOut)
xout.setStreamName("sync_out")
sync.out.link(xout.input)
return pipeline
def colorize_depth_for_display(depth_frame):
depth_normalized = cv2.normalize(depth_frame, None, 0, 255, cv2.NORM_MINMAX)
depth_colored = cv2.applyColorMap(np.uint8(depth_normalized), cv2.COLORMAP_JET)
return depth_colored
if __name__ == '__main__':
pipeline = create_pipeline()
with dai.Device(pipeline) as device:
q_sync = device.getOutputQueue(name="sync_out", maxSize=4, blocking=False)
frame_counter = 0
save_dir = None
while True:
msgGrp = q_sync.get()
frames = {}
for name, msg in msgGrp:
frames[name] = msg.getCvFrame()
if len(frames) == 3: # We expect 3 frames (ToF, left RGB, right RGB)
# Get raw frames
tof_depth = frames['tof']
left_rgb = frames['left']
right_rgb = frames['right']
# Display frames (colorize depth only for display)
cv2.imshow("ToF Depth", colorize_depth_for_display(tof_depth))
cv2.imshow("Left RGB", left_rgb)
cv2.imshow("Right RGB", right_rgb)
key = cv2.waitKey(1)
if key == ord('q'):
break
elif key == ord('c'):
# Create new directory for saving if it doesn't exist
if save_dir is None:
timestamp = time.strftime("%Y%m%d_%H%M%S")
save_dir = os.path.join("data", timestamp)
os.makedirs(save_dir, exist_ok=True)
print(f"Created directory: {save_dir}")
# Save raw frames
tof_filename = os.path.join(save_dir, f"tof_{frame_counter:06d}.npy")
left_filename = os.path.join(save_dir, f"left_{frame_counter:06d}.png")
right_filename = os.path.join(save_dir, f"right_{frame_counter:06d}.png")
# Save ToF as raw numpy array, RGB images as PNG
np.save(tof_filename, tof_depth)
cv2.imwrite(left_filename, left_rgb)
cv2.imwrite(right_filename, right_rgb)
print(f"Saved frame {frame_counter}")
frame_counter += 1
cv2.destroyAllWindows()
print("Program finished")Thanks,
Jaka
Dear @jakaskerl ,
Thank you very much for your help.
Here are the files I got in a bigger room:
Link to Photofiles
As mentioned I am trying to achive a sendout of color, depth and pointcloud with as less latency as possible and as good Rgb Quality as possible by keeping my osc rgb and tof camera control I implemented in my last script.
Link to last script
Here in parallel unfortunately I had to pay the price, becaus my production is already running => my approach with Orbbec Femto Mega.
To clear out what I am trying. If quality and latentcy is ok and the alignment as prefect as those from the Femto Mega I would slice the pointcloud as in the video example and use the combination of depthmap and sliced pointcloud as alpha channel to cutout the parts that have to become visible.
Link to my Femto Mega approach
Thank you again very much for your help.
As an information:
Luxonis could become serious competitor to the Femto Mega or even the Femto Mega 1 with the 12MP Tof sensor or even with the Oak 4 Pro (with ToF)
because Femtos Color Cameras have not such an easy live control + not such a good quality with their color cameras.
They are only more practical at the moment because of their direct connection through the Ip adress in Touchdesigner and their pretty exact point cloud.
For visual arts a good Oak 4 Pro with Tof option that is well aligned could beat Orbbec in speed (2.5 Gbs) and quality, especially if you bind in another ToF sensor.
Thats why I am not shure, that if I don't achieve the quality with my custom build, if it would be better to exchange my sensor with another camera from OaK.
Anyway I am very thankul for all your help and please be honest If you think if I could reach Femto Mega with my custom Oak Tof or if I should redirect my money probably to something new like the Oak 4 Pro (wide) or the interesting Oak Thermal sensor.
Hi BonkoKaradjov
Team has had a chance to look at your scripts.
Issue was that your calibration was done using rotated TOF frame not RGB, so the alignment has to also be done for rotated TOF and not rotated RGB frame, otherwise the rot matrix is incorrect.
Code for properly aligning the streams:
import os
import pathlib
import numpy as np
import cv2
import depthai as dai
from numba import jit, prange
@jit(nopython=True, parallel=True)
def reprojection(depth_image, depth_camera_intrinsics, camera_extrinsics, color_camera_intrinsics, depth_image_show = None):
height = len(depth_image)
width = len(depth_image[0])
if depth_image_show is not None:
image = np.zeros((height, width), np.uint8)
else:
image = np.zeros((height, width), np.uint16)
if(camera_extrinsics[0][3] > 0):
sign = 1
else:
sign = -1
for i in prange(0, height):
for j in prange(0, width):
if sign == 1:
# Reverse the order of the pixels
j = width - j - 1
d = depth_image[i][j]
if(d == 0):
continue
# Convert pixel to 3d point
x = (j - depth_camera_intrinsics[0][2]) * d / depth_camera_intrinsics[0][0]
y = (i - depth_camera_intrinsics[1][2]) * d / depth_camera_intrinsics[1][1]
z = d
# Move the point to the camera frame
x1 = camera_extrinsics[0][0] * x + camera_extrinsics[0][1] * y + camera_extrinsics[0][2] * z + camera_extrinsics[0][3]
y1 = camera_extrinsics[1][0] * x + camera_extrinsics[1][1] * y + camera_extrinsics[1][2] * z + camera_extrinsics[1][3]
z1 = camera_extrinsics[2][0] * x + camera_extrinsics[2][1] * y + camera_extrinsics[2][2] * z + camera_extrinsics[2][3]
u = color_camera_intrinsics[0][0] * (x1 / z1) + color_camera_intrinsics[0][2]
v = color_camera_intrinsics[1][1] * (y1 / z1) + color_camera_intrinsics[1][2]
int_u = round(u)
int_v = round(v)
if int_u >= 0 and int_u < (len(image[0]) - 1) and int_v >= 0 and int_v < len(image):
if depth_image_show is not None:
image[int_v][int_u] = depth_image_show[i][j][0]
image[int_v][int_u + sign] = depth_image_show[i][j][0]
else:
image[int_v][int_u] = z1
image[int_v][int_u + sign] = z1
return image
def colorizeDepth(frameDepth):
invalidMask = frameDepth == 0
# Log the depth, minDepth and maxDepth
try:
minDepth = np.percentile(frameDepth[frameDepth != 0], 3)
maxDepth = np.percentile(frameDepth[frameDepth != 0], 95)
logDepth = np.log(frameDepth, where=frameDepth != 0)
logMinDepth = np.log(minDepth)
logMaxDepth = np.log(maxDepth)
np.nan_to_num(logDepth, copy=False, nan=logMinDepth)
# Clip the values to be in the 0-255 range
logDepth = np.clip(logDepth, logMinDepth, logMaxDepth)
# Interpolate only valid logDepth values, setting the rest based on the mask
depthFrameColor = np.interp(logDepth, (logMinDepth, logMaxDepth), (0, 255))
depthFrameColor = np.nan_to_num(depthFrameColor)
depthFrameColor = depthFrameColor.astype(np.uint8)
depthFrameColor = cv2.applyColorMap(depthFrameColor, cv2.COLORMAP_JET)
# Set invalid depth pixels to black
depthFrameColor[invalidMask] = 0
except IndexError:
# Frame is likely empty
depthFrameColor = np.zeros((frameDepth.shape[0], frameDepth.shape[1], 3), dtype=np.uint8)
except Exception as e:
raise e
return depthFrameColor
def get_Calibration_Data(calibData: dai.CalibrationHandler, TOF_SOCKET, ALIGN_SOCKET, depthSize = (640, 480), rgbSize = (1920, 1080)):
M1 = np.array(calibData.getCameraIntrinsics(TOF_SOCKET, *depthSize))
D1 = np.array(calibData.getDistortionCoefficients(TOF_SOCKET))
M2 = np.array(calibData.getCameraIntrinsics(ALIGN_SOCKET, *rgbSize))
D2 = np.array(calibData.getDistortionCoefficients(ALIGN_SOCKET))
T = (
np.array(calibData.getCameraTranslationVector(TOF_SOCKET, ALIGN_SOCKET, False))
* 10
) # to mm for matching the depth
R = np.array(calibData.getCameraExtrinsics(TOF_SOCKET, ALIGN_SOCKET, False))[
0:3, 0:3
]
TARGET_MATRIX = M1
return M1, D1, M2, D2, TARGET_MATRIX, R, T
def getAlignedDepth(frameDepth, TARGET_MATRIX, M2, R, T, rgbSize = (1920, 1080)):
R_180 = np.array([[-1., 0., 0.],
[0., -1., 0.],
[0., 0., 1.]])
combinedExtrinsics = np.eye(4)
combinedExtrinsics[0:3, 0:3] = R
combinedExtrinsics[0:3, 3] = T
depthAligned = reprojection(frameDepth, TARGET_MATRIX, combinedExtrinsics, TARGET_MATRIX)
mapX, mapY = cv2.initUndistortRectifyMap(TARGET_MATRIX, None, np.eye(3), M2, rgbSize, cv2.CV_32FC1)
outputAligned = cv2.remap(depthAligned, mapX, mapY, cv2.INTER_NEAREST)
return outputAligned
import argparse
parser = argparse.ArgumentParser()
parser.add_argument("-path", help="Path to the dataset, were all files are.")
args = parser.parse_args()
### Write down, which sockets you wanna align to ####
ALIGN_SOCKET = dai.CameraBoardSocket.CAM_C
TOF_SOCKET = dai.CameraBoardSocket.CAM_A
path = args.path
calibData = dai.CalibrationHandler(path + "/calib.json")
for files in os.listdir(path):
### LOADING ALL images which are used ###
if files.split(".")[-1] == "json":
continue
index = files.split(".")[0].split("_")[1]
depth = np.load(path + f"/tof_{index}.npy")
left = cv2.imread(path + f"/left_{index}.png")
right = cv2.imread(path + f"/right_{index}.png")
### LOGIC ####
# Logic is following:
# Since device was calibrated with the rotated depth, the depth frame MUST be rotated before aligned
depth = cv2.rotate(depth, cv2.ROTATE_180)
# From data you have, get all needed matrices
M1, D1, M2, D2, TARGET_MATRIX, R, T = get_Calibration_Data(calibData, TOF_SOCKET, ALIGN_SOCKET, (depth.shape[1], depth.shape[0]), (right.shape[1], right.shape[0]))
#Align rotated depth to the rotated RGB image
alignedDepth = getAlignedDepth(depth, TARGET_MATRIX, M2, R, T, (right.shape[1], right.shape[0]))
#Rotate back both depth and RGB
right = cv2.rotate(right, cv2.ROTATE_180)
alignedDepth = cv2.rotate(alignedDepth, cv2.ROTATE_180)
blended = cv2.addWeighted(right, 0.5, colorizeDepth(alignedDepth), 0.5, 0)
cv2.imshow("Blended", blended)
cv2.waitKey(0)Thanks,
Jaka
Dear Jaka, thank you very much for your help,
To be honest, I lost the job, the camera was planned in, and all the money and Luxonis does not want take back this custom build brick, even ignores my last email..but ok.
At least I payed extra 300 € for a calculated sensor I asked for and received a ......
But I don't want to feel so shitty that I posses a 800 € brick working as the most expensive 12MP camera ever selled!!
I am afraid I will never calibrate that sensor right. Your script, if I understood it correctly needs the calibration pictures with a charuco board first, which I tried.....
After following the calibration instructions here:
LINK Luxonis Website
I first completely bricked my pycharm virtual environment. Now it either says "no sensor detected" or does not recognize certain parts such as the FPPN_ENABLE etc. in none of my written scripts even the examples from Luxonis don't work... so now the sensor is on niveau even more expensive paperweight.
My calibration problems:
A Forum tells that the aruco Api had changed for certain versions.
So I should rewrite the Luxonis calibrate.py ???
I am afraid that if I solve this issue the calibration will not start again and new errors will appear.
Is it right from Luxonis that when receiving a wrong calibrated sensor and I payed for that I have to invest a huge amount of time in finding out how to calibrate it myself and living with the problem that all future upcoming scripts for it will not work and cost me more time and therefore paying extra 300 €?
I was at least hoping that they will take back the order and send me a standard, well calibrated version (that has the same components...I even can not think why they think this sensor is unsellable again - it has the same components ) + give me back my 300 € for the custom build.
To be honest: Buying the 12MP camera from somewhere else and soldering it to the board by a friend would have been much cheaper...and was not the reason why I send the sensor back to Luxonis for the "custom build"
I was very thankful in the beginning with their great support but now I am seriously disappointed, now that I know that the support was not aware of the fact that my sensor was wrongly calibrated and I wasted soo much time, my time and the time of the support.
Which I am very sorry for.
I wish I could have used the support and this forum for more productive questions than, "please help me how to use my brocken device and repair it with me......"
I will of course have to write this into my test review for ToF Sensors for Artists to never send Luxonis a sensor for custom build, even if it has the exact same components as an industrial build from them if you work on a job, even for a job in the future.
But thank you very, very much for your help.
Any suggestions @jakaskerl @erik what is wrong to try at least myself to calibrate that sensor anyhow?
I was getting errors before downgrading to numpy1, for whatever reason there is a requirements installer.
I am only trying to get the pictures to use the script from @jakaskerl above to get the correct calibration.
If I write it as the first example I get this beautiful error:
BonkoKaradjov
Why are you trying to recalibrate the device? Everything should be calibrated already.
Thanks,
Jaka
jakaskerl Thank you very much for your help.
I maybe misunderstood the workflow.
I get errors because your script is searching for a calib.json and the photo files.
That brought me to the misunderstanding that I first have to get the images from a charuco board from different angles so that this script makes the alignment?
So for example: Could I use this script of mine and modify it?
@jakaskerl Look at my final approach with the Femto Mega and Touchdesigner:
If I could get Color Stream + Depth Stream + Pointcloud Stream through NDI or Spout perfectly aligned out
in combination with my scrip that controls everything through OSC this could beat the Femto Mega immediately because Femto Megas image quality is way less worse than the image quality from your 12MP ToF sensor.
Here what I get in Touchdesigner from the Orbbec Top through network:
[https://gyazo.com/96490f9625ddef7833c3e4dbb4352896](https://)
BonkoKaradjov
Camera should already calibrated, you shouldn't need further calibration.
You can use your script, but make sure to use alignment from https://discuss.luxonis.com/d/5692-how-to-send-out-32bit-tof-depthmap-through-spout/11 in order to achieve proper alignment.
LMK what the main issue is with NDIlib if you run into any.
Thanks,
Jaka
Dear @jakaskerl , Dear @erik , Thank you very much for all your help.
I want to apologize for my emotional words above. Somehow I was totaly confused and misunderstood completely and thought that the sensor not had been calibrated and was upset that because my lack of knowledge (so my fault) I could not prepare the sensor for the one arts job I planned.
Now the custom Oak ToF is on stack and I have time to research its possibilities and hope to reach the goal one day.
I just wanted to say thank you for all your help and patience.
I really appreciate what Luxonis support does and hope that more people could know about your company and buy your sensors.
With kind regards
Bonko
My uncle used to say if you chase chickens, you get feathers. If you chase strategy, you get crypto. The thing about gaming on blockchain-powered portals is that you need speed, stealth, and a seriously smart approach. Sites like https://bgpochivnidni.eu/ list platforms like CoinCasino, where crypto deposits are instant, and you’ve got thousands of titles to test your tactical acumen Stick to games with transparent odds, watch your bankroll like it’s your last pizza slice, and skip the emotional drama. Math is your best buddy in this business.
