@linus777 , the instructions you referenced were made for VOXL1, which had all cameras connected to the same i2c bus and slave address conflicts were not allowed. On VOXL2, each camera port has its own i2c bus and there is no possibility of a conflict when using several cameras with the same i2c slave addresses. Is your question related to usage on VOXL1 or VOXL2? Alex

@tom said in Can someone tell me which camera is used in this video?: Flir Boson connected over USB Thank you!

@Jetson-Nano , Can you please double check your /etc/modalai/voxl-camera-server.conf to make sure that the tracking camera has "en_raw_preview" set to true? It is hard to tell whether there is haze in the image (the tracking camera has 640x480 resolution and if you zoom in, it may appear that the details are blurry). If you have another tracking camera to compare to, please try. Also please try looking at the image from voxl-portal (original tracking image and overlay image). Alex

@Moderator said in Thermal Camera to VOXL2: @H-AL first you need to determine if the sensor's electrical interface is compatible. What interface exists on the sensor? Can you find an appropriate electrical connection on VOXL 2? Then, you need to find if there is software compatibility, does the driver current exist in ModalAI's reference code? Or, is there an implementation you can leverage for your work? What thermal camera do you recommend using?

@Alex-Kushleyev Alright, thanks for the info

@Alex-Kushleyev Hi Alex, awesome! - and that's a good question. We don't know necessarily where the text will be in the image beforehand, so do we have to look for it. On the other hand, looking for text and reading it are two different stages - so we are planning to try a pipeline where we first find text at a lower resolution/lower quality and then process selected crops with more resolution and/or processing. So, in short, debayering only selected parts of the image is something that we are interested in, but we only know what regions to process at runtime, not before we look at the image. If debayering selected regions is possible that's something we are definitely interested in. Thanks!

@H-AL , are you using voxl-camera-server to produce h264 / h265 stream or voxl-streamer, which is capable of accepting uncompressed frames and producing h264/h265 stream. voxl-streamer is also capable of taking a compressed stream from voxl-camera-server and exposing it via RTSP. Please clarify how you are setting up your streaming.. Thanks! Alex

@Alex-Kushleyev Thanks! This is helpful. Definitely easier to overheat these than the last gen so good to have the specs on it.

@Gyeongmin-Kim We don't have any great insight at this point. Those sensors are new. It looks like you are seeing some ghosting. You could try different frame rates or exposure times to try and minimize the artifacts

@MattK @Moderator If the Starling V2 Max doesn't support a DFS server due to IMX412 being rolling shutter cameras and no TOF, then how is the 3D map generated? Since in my understanding, to generate a map you need depth values/ point cloud data. If the depth estimates are being obtained by some other means then why can't the same point cloud be made available for VOA in vision-hub server?

@Jeremy-Frederick, that is great! Luckily (or by design ) the camera connector reversal is typically not fatal, but DO NOT push your luck, please . Technically, if damage occurs due to incorrect insertion of the camera connectors, this will void the warranty for the main board. Always verify connector orientation when changing camera configuration. Alex

@Gary-Holmgren , camera server does not support rotating images by 90 degrees (or multiples of 90). You could do this yourself in your image processing application. If you mount the cameras rotated by 90 degrees (while keeping the baseline along the left-right direction of the drone), you would need to modify your extrinsics to account for this rotation. If you perform a stereo camera calibration in this rotated configuration, then the output of the stereo calibration should give you the baseline result between the cameras in the Y direction in the camera frame (as opposed to X direction in normal case). If you have a question about a specific configuration, please make a drawing that shows reference frames of the vehicle, imu, and both cameras and I can help you double check the extrinsics. Alex

@dlee we don't have those benchmarks yet, sorry. The IRS2975C based TOF should go 1-2m further than the old one

@Alex-Kushleyev Oh, very fascinating. Thank you for the update! I'd actually somewhat given up on DFS and started testing your new ToF beta product, but when I can I'll try boot my DFS rig back up and see if the fix helps. Edit: Nevermind, I just checked and my DFS rig is on SDK v1.1.2, so this fix shouldn't be applicable. I might still try it out just in case, though.

@Alex-Kushleyev That worked, thank you! And yes, it is the rolling shutter effect, I believe it was just more noticeable with the color change. Everything seems to be working now!

if CPU was causing the issue, then you would see slower FPS, not degraded performance. This sounds like the actual TOF sensor is heating up and at some point starts throttling itself by reducing the emitted power, which helps cool it down. You could test it by selectively cooling TOF sensor (and not voxl) and see if performance goes back up.

@Aaky, yes if you change the camera configuration, just follow the same process: copy correct sensorbin (with correct slot id) to /usr/lib/camera/ as previously discussed confirm the camera is detected using voxl-camera-server -l update camera config xml with correct camera ids - the camera id’s will start with zero and increment by 1 for each camera. The camera id assignment happens in order of the camera slot id. Slot id is mapped to the physical connector. You may want to remove unused sensorbins from /usr/lib/camera just to avoid any confusion. You can always get them back from /usr/share/modalai/chi-cdk

@Darshit-Desai , yes, feel free to modify the code as you need for right now for your own testing. Once i have a chance, i will make the mode selection (long / short) and FPS more flexible. I hope this unblocked your work and you can move forward. Alex

@Alex-Kushleyev , I think that is plenty for now. We will order a couple and test them. Thanks a lot!