Hi, So I saw the [Depth Crop Control](https://docs.luxonis.com/software/depthai/examples/depth_crop_control/) example and I was working on it to update it to crop control using the ISP image. However I get the following output image [upl-image-preview url=https://discuss.luxonis.com/assets/files/2024-08-16/1723821588-449884-screenshot-2024-08-16-095339.png] 1. I am getting that strange image area below the cropped region as you can see. Why is that happening any idea? 2. I understand that ISP outputs YUV420 planar image. When I am printing the shape of the output cropped ISP image from the ImageManip node it says its (2736, 2434). Does this mean its a single channel image like a grayscale image? When I print the shape of an RGB frame I do see 3 channels in the shape as (3, W, H). I look forward to hearing from you soon. Thanks in advance. Thanks & Regards Yishu

@"yishu_corpex"#2472 1. Can you add some minimal reproducible code? It's hard to say what is going on here. 2. The YUV420 is stored as an array of first Y (luma) values and then subsampled U and V (chroma) values. It doesn't have channels. If you do a getData() on 1080p image you will see it has a size of `2,073,600 (1920*1080 Y) + 518,400 (540*960 U) + 518,400 (540*960 V) = 3,110,400 pixels`. It's all in a single channel. The shape of that would be (1620, 1920). When converting to BGR via `getCvFrame()` the U and V layers are upsampled to Y dimensions and combined together to form BGR (3 channel data). Thanks, Jaka

ISP Crop Control

yishu_corpex

Hi,

So I saw the Depth Crop Control example and I was working on it to update it to crop control using the ISP image.

However I get the following output image

I am getting that strange image area below the cropped region as you can see. Why is that happening any idea?
I understand that ISP outputs YUV420 planar image. When I am printing the shape of the output cropped ISP image from the ImageManip node it says its (2736, 2434). Does this mean its a single channel image like a grayscale image? When I print the shape of an RGB frame I do see 3 channels in the shape as (3, W, H).

I look forward to hearing from you soon. Thanks in advance.

Thanks & Regards
Yishu

jakaskerl

@yishu_corpex

Can you add some minimal reproducible code? It's hard to say what is going on here.
The YUV420 is stored as an array of first Y (luma) values and then subsampled U and V (chroma) values. It doesn't have channels. If you do a getData() on 1080p image you will see it has a size of 2,073,600 (1920*1080 Y) + 518,400 (540*960 U) + 518,400 (540*960 V) = 3,110,400 pixels. It's all in a single channel. The shape of that would be (1620, 1920).
When converting to BGR via getCvFrame() the U and V layers are upsampled to Y dimensions and combined together to form BGR (3 channel data).

Thanks,
Jaka