search

DeyvidKochanov-TomTom / kprnet

MIT License
66 stars 12 forks source link

some question about the projuection #18

Closed degagex closed 2 years ago

degagex commented 2 years ago

Hi,I'm reading you code now,in your projection,I change the threshold (you set it 0.2 and 0.8) to any figure between 0 and 1, but the value of new_raw is always 63,I wonder know why it comes that,thanks a lot!

DeyvidKochanov-TomTom commented 2 years ago

Hi, in this issue I tried to explain what the values 0.2 and 0.8 are. I don't know why you would need to change them but I guess almost anything will work e.g. setting them to 0.1 and 0.9 would also be okay. I hope this helps!

degagex commented 2 years ago

Thanks!Tom. HHH,maybe I didn't express my opion clearly,I mean I can't understand that why do this "new_raw = np.nonzero((proj_x[1:] < 0.2) * (proj_x[:-1] > 0.8))[0] + 1" ,the value of new_raw is 63 exactly?

DeyvidKochanov-TomTom commented 2 years ago

I am not sure I understand the question if you want to understand the 0.2 and 0.8 values I suggest you look at the other issue. I don't know why for you new_raw always becomes 63, it should be an array which contains the indexes of all points which start a new line in the lidar scan.

degagex commented 2 years ago

OK OK，I get it ，thanks alot！

---Original--- From: @.> Date: Tue, Jan 25, 2022 20:33 PM To: @.>; Cc: @.**@.>; Subject: Re: [DeyvidKochanov-TomTom/kprnet] some question about the projuection (Issue #18)

I am not sure I understand the question if you want to understand the 0.2 and 0.8 values I suggest you look at the other issue. I don't know why for you new_raw always becomes 63, it should be an array which contains the indexes of all points which start a new line in the lidar scan.

— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.Message ID: @.***>

degagex commented 2 years ago

I don't know why for you new_raw always becomes 63, it should be an array which contains the indexes of all points which start a new line in the lidar scan. I guess because I use SemanticKITTI dataset,it gets data vy velodyne 64？I am just a beginner,so maybe my questions are funny😂

---Original--- From: @.> Date: Tue, Jan 25, 2022 20:33 PM To: @.>; Cc: @.**@.>; Subject: Re: [DeyvidKochanov-TomTom/kprnet] some question about the projuection (Issue #18)

I am not sure I understand the question if you want to understand the 0.2 and 0.8 values I suggest you look at the other issue. I don't know why for you new_raw always becomes 63, it should be an array which contains the indexes of all points which start a new line in the lidar scan.

— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.Message ID: @.***>

DeyvidKochanov-TomTom commented 2 years ago

The code should work with semantic kitti so if you can give an example by sequence id and point cloud id I can look at it. But also are you doing any kind of augmentation to the point cloud beforehand like random flipping? That would break the projection function.

degagex commented 2 years ago

I don't do any change of the code,and I use KITTI sequce00 and 000000.bin,iy contains 124668 points.

---Original--- From: @.> Date: Tue, Jan 25, 2022 20:56 PM To: @.>; Cc: @.**@.>; Subject: Re: [DeyvidKochanov-TomTom/kprnet] some question about theprojuection (Issue #18)

The code should work with semantic kitti so if you can give an example by sequence id and point cloud id I can look at it. But also are you doing any kind of augmentation to the point cloud beforehand like random flipping? That would break the projection function.

— Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you authored the thread.Message ID: @.***>

DeyvidKochanov-TomTom commented 2 years ago

Hi, I tried the scan but it seems okay. Could you put together a minimal example to reproduce your problem? Something similar to this: https://gist.github.com/DeyvidKochanov-TomTom/9551bb36bbf87f9c7437f70a5ed1173c

degagex commented 2 years ago

Hi, I know your code now, this code "new_raw = np.nonzero((proj_x[1:] < 0.2) * (proj_x[:-1] > 0.8))[0] + 1" is used to find the points which begin the new line, 0.8 and 0.2 are not the unique threshold, thanks a lot, you are a nice man 😉!