Arducam MIPI_Camera got green frame

lvbin · January 31, 2021, 4:56pm

Hello,

Sorry for our document mislead you. Only when you use raspistill tool. it supports AE and AWB function due to it has RPI hardware ISP support. Using our MIPI_Camera driver, due to it does not have ISP support, so we just have software AE and AEB function and the result is not very well. Currently our ISP is in the theoretical testing stage and made great process.

bolteon · January 31, 2021, 10:44pm

Thank you for replying Bin.

Do you have a direct email? I have some more questions that I would like to keep off the main forums; I also contacted regarding a custom order for design of a camera solution, have not heard back from anyone in over a month…

lvbin · February 1, 2021, 9:18pm

You are welcome. We can chat through skype ID: “[email protected]”

or send email to our [email protected]. I will try my best to help you.

lucabu · March 8, 2021, 12:23am

Any news about AWB & AE support? when will they be available?

right now we all have a pretty much useless kit on our desks…

Thanks

lvbin · March 12, 2021, 7:11pm

Hi, Our ISP is still developing. Please waiting .

alo · May 24, 2021, 9:25am

Greetings,

I have the same problem. Not only is the image green but the frame rate is wrong and the calibration does not work.

I’m using: Raspberry Pi 4 4GB

I’ve purchased: Arducam 8MP Synchronized Stereo Camera Bundle Kit for Raspberry Pi

SKU: B0195S8MP

Your tutorial video shows the product with no green image. Also, it presents the 5MP and the 8MP camera for sale in the description. So it is perfectly acceptable to assume that this tutorial is for both 5MP and 8MP for the Raspberry Pi. The video: https://www.youtube.com/watch?v=eBZm40z7E8Y&ab_channel=Arducam

I’ve also read every webpage found in: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/sync-stereo-camera-hat-user-manual/#ftoc-heading-9 . Here states an introduction for the stereo camera pair (Sync Stereo Camera HAT User Manual). I’ve read and did everything that was instructed (except points 5.4.3 and 5.4.4 because they do not work for this stereo pair for some reason). Both 5MP and 8MP were introduced there.

Then, I followed your instructions at: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/opencv-and-depth-map-on-arducam-stereo-camera-hat-tutorial/ . Running the first program (python3 1_test.py) shows a green image. If the reason is because of ISP or AWB and the algorithm is “suppost” to show a green image then why do you present a “regular” image in the youtube video? Does this not affect calibration??

While trying to fix the green image problem I added to the code:

camera.software_auto_exposure(enable = True) #Added this to adjust light

and camera.software_auto_exposure(enable = True) #Added this to adjust light

It gives a red/pink tone to the image but it corrects itself somewhat appropriately.

Not only that but the recorded frame rate is not what you present in the websites mentioned above. For example, for the 8MP and mode: 8, width: 2560, height: 720, 38fps. When I save the image with 1_test1.py it gives me an average of 17fps in the best cases and 12fps in the worst. Not even close to 38fps.

Lastly, after running 1_test.py, 2_chess_cycle.py, 3_pairs_cut.py and 4_calibration.py the algorithm can not find the pattern you provide. I tried changing the pattern parameters with no success.

Summing up the problems:

The image from Raspistill and Raspivid is much better when compared with what is presented by the Arducam software. Both the color and fps have drastic differences.
Points 5.4.3 and 5.4.4 of the initial tutorial do not work.
Green image problem
Calibration can not be performed because the pattern is not recognized.

It is not correct to present a product with results that are different from reality. It is unacceptable that a company sells a product that does not work. This is clearly a common topic in this product. However, before accusing anyone of anything I first recognize that I could be doing something wrong with some or all of these topics. Therefore, I’ve sent and email nearly two weeks ago but got no answer. Even creating an account in this forum was problematic and made me use two emails! Please tell me a good way to communicate with you to solve this as, so far, I have a something that clearly does not work. I’m available any day at any time and I more than willing to report problems/bugs to help you develop this amazing product.

Thank you for your time.

Best regards.

alo · May 24, 2021, 9:32am

Greetings,

I have the same problem. Not only is the image green but the frame rate is wrong and the calibration does not work.

I’m using: Raspberry Pi 4 4GB

I’ve purchased: Arducam 8MP Synchronized Stereo Camera Bundle Kit for Raspberry Pi

SKU: B0195S8MP

Your tutorial video shows the product with no green image. Also, it presents the 5MP and the 8MP camera for sale in the description. So it is perfectly acceptable to assume that this tutorial is for both 5MP and 8MP for the Raspberry Pi. The video: https://www.youtube.com/watch?v=eBZm40z7E8Y&ab_channel=Arducam

I’ve also read every webpage found in: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/sync-stereo-camera-hat-user-manual/#ftoc-heading-9 . Here states an introduction for the stereo camera pair (Sync Stereo Camera HAT User Manual). I’ve read and did everything that was instructed (except points 5.4.3 and 5.4.4 because they do not work for this stereo pair for some reason). Both 5MP and 8MP were introduced there.

Then, I followed your instructions at: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/opencv-and-depth-map-on-arducam-stereo-camera-hat-tutorial/ . Running the first program (python3 1_test.py) shows a green image. If the reason is because of ISP or AWB and the algorithm is “suppost” to show a green image then why do you present a “regular” image in the youtube video? Does this not affect calibration??

While trying to fix the green image problem I added to the code:

camera.software_auto_exposure(enable = True) #Added this to adjust light

and camera.software_auto_exposure(enable = True) #Added this to adjust light

It gives a red/pink tone to the image but it corrects itself somewhat appropriately.

Not only that but the recorded frame rate is not what you present in the websites mentioned above. For example, for the 8MP and mode: 8, width: 2560, height: 720, 38fps. When I save the image with 1_test1.py it gives me an average of 17fps in the best cases and 12fps in the worst. Not even close to 38fps.

Lastly, after running 1_test.py, 2_chess_cycle.py, 3_pairs_cut.py and 4_calibration.py the algorithm can not find the pattern you provide. I tried changing the pattern parameters with no success.

Summing up the problems:

The image from Raspistill and Raspivid is much better when compared with what is presented by the Arducam software. Both the color and fps have drastic differences.
Points 5.4.3 and 5.4.4 of the initial tutorial do not work.
Green image problem
Calibration can not be performed because the pattern is not recognized.

It is not correct to present a product with results that are different from reality. It is unacceptable that a company sells a product that does not work. This is clearly a common topic in this product. Before accusing anyone of anything I first recognize that I could be doing something wrong with some or in all of these topics. Therefore, I’ve sent and email nearly two weeks ago but got no answer. Even creating an account in this forum was problematic and made me use two emails! Please tell me a good way to communicate with you to solve this as, so far, I have a something that clearly does not work. I’m available any day at any time and I’m more than willings to report problems/bugs to help you develop this amazing product.

Thank you for your time.

Best regards.

alo · May 24, 2021, 9:33am

Greetings,

I have the same problem. Not only is the image green but the frame rate is wrong and the calibration does not work.

I’m using: Raspberry Pi 4 4GB

I’ve purchased: Arducam 8MP Synchronized Stereo Camera Bundle Kit for Raspberry Pi

SKU: B0195S8MP

Your tutorial video shows the product with no green image. Also, it presents the 5MP and the 8MP camera for sale in the description. So it is perfectly acceptable to assume that this tutorial is for both 5MP and 8MP for the Raspberry Pi. The video: https://www.youtube.com/watch?v=eBZm40z7E8Y&ab_channel=Arducam

I’ve also read every webpage found in: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/sync-stereo-camera-hat-user-manual/#ftoc-heading-9 . Here states an introduction for the stereo camera pair (Sync Stereo Camera HAT User Manual). I’ve read and did everything that was instructed (except points 5.4.3 and 5.4.4 because they do not work for this stereo pair for some reason). Both 5MP and 8MP were introduced there.

Then, I followed your instructions at: https://www.arducam.com/docs/cameras-for-raspberry-pi/synchronized-stereo-camera-hat/opencv-and-depth-map-on-arducam-stereo-camera-hat-tutorial/ . Running the first program (python3 1_test.py) shows a green image. If the reason is because of ISP or AWB and the algorithm is “suppost” to show a green image then why do you present a “regular” image in the youtube video? Does this not affect calibration??

While trying to fix the green image problem I added to the code:

camera.software_auto_exposure(enable = True) #Added this to adjust light

and camera.software_auto_exposure(enable = True) #Added this to adjust light

It gives a red/pink tone to the image but it corrects itself somewhat appropriately.

Not only that but the recorded frame rate is not what you present in the websites mentioned above. For example, for the 8MP and mode: 8, width: 2560, height: 720, 38fps. When I save the image with 1_test1.py it gives me an average of 17fps in the best cases and 12fps in the worst. Not even close to 38fps.

Lastly, after running 1_test.py, 2_chess_cycle.py, 3_pairs_cut.py and 4_calibration.py the algorithm can not find the pattern you provide. I tried changing the pattern parameters with no success .

Summing up the problems:

The image from Raspistill and Raspivid is much better when compared with what is presented by the Arducam software. Both the color and fps have drastic differences.
Points 5.4.3 and 5.4.4 of the initial tutorial do not work.
Green image problem
Calibration can not be performed because the pattern is not recognized.

It is not correct to present a product with results that are different from reality. It is unacceptable that a company sells a product that does not work. This is clearly a common topic in this product. Before accusing anyone of anything I first recognize that I could be doing something wrong with some or in all of these topics. Therefore, I’ve sent and email nearly two weeks ago but got no answer. Even creating an account in this forum was problematic and made me use two emails! Please tell me a good way to communicate with you to solve this as, so far, I have a something that clearly does not work. I’m available any day at any time and I’m more than willings to report problems/bugs to help you develop this amazing product.

Thank you for your time.

Best regards.

alo · May 26, 2021, 1:12am

Hello there!
I’m going to reply to myself here. I spent all day trying new things with this program and these are the conclusions I took:

In the previous post I mentioned the same line of code twice. Here are both:
camera.software_auto_exposure(enable = True) #Added this to adjust light
camera.software_auto_white_balance(enable = True) #Added this to adjust color
First major difference from yesterday is the fact that I finally realized that the rows and columns do not mean the actual rows and columns of the chessboard but the “Number of inside corners in the chessboard’s rows.” so their pattern is a 6x9 and not a 7x10 as I wrongly assumed. This finally meant I could perform the Calibration (because the pattern was not recognized before). I saw at least two other people with this problem in forums and websites, hope this helps.
Using the lines of code I mentioned earlier sometimes improves calibration a little (I think it is because the image is brighter) but could be coincidental (and therefore using “green mode” could present the same results). However the image is closer to reality. Also, using these lines of code slows the camera display process a bit. For example for mode 8 (which is suppost to do 38fps) shows around 15fps in “green mode” (without the lines of code) and around 13 with the code.
When taking photos for calibration (2_chess_cycle.py) the pattern can’t be too far away for the camera otherwise the pattern will sometimes not be correctly recognized (even though it is later accepted by function 4 (4_calibration.py). I think this is the main cause for people complaining about the calibration having some weird results. Taking close up pictures of the pattern, rotating it in all 3 axis, gave me a good calibration result.
The modes have some weird differences between each other concerning FOV. I didn’t measure the angles but to sum up:
5.1 In the horizontal plane the angle of view was (mode) 7=11=12 > 8=10 > 9
5.2 In the vertical plane the angle of view was (mode) 7=12 > 11=10 > 8 > 9
Note that 8>9 for example means that mode 8 showed more in the vertical direction that mode 9 and that 7=11 means that mode 7 was VERY similar to mode 11 (not necessarily the same, as I didn’t measure the exact angles)
Ultimately Arducam did a very good job making the program. However, there are some weird results that I still can’t explain (or find explanation of).
Again, I would like to say that I’m available to explain everything I did and would appreciate any help/criticism from both consumers and Arducam.
Thank you for your time.
Best regards.

maw · September 23, 2021, 5:07pm

Hi there

I recently purchased the 12MP IMX477 Synchronized Stereo Camera Bundle Kit and experienced the same issue (green image) using the MIPI_Camera driver .
Using a RaspberryPi 4B, I followed this guideline step by step. While the raspistill command works perfectly, the arducamstill command with mode 1 (8112×3040 resolution) results in a weird image where I get something in the left half in a greenish look, while the right one stays black:

Is there something I can do about this?

Thanks in advance!