search

EmaroLab / PDT

Detection and Tracking of Pallets using a Faster R-CNN based on a 2D LRF
42 stars 15 forks source link

The content in GrayIm and in raw laserscanner data with the same image number doesn't match #1

Closed tin1254 closed 4 years ago

tin1254 commented 4 years ago

I read and convert the laserscanner data and from txt and compare it with the Graylm. Starting from GrayIm, the image contents in the laserscanner data and Graylm don't match starting from scan 131 Screenshot from 2020-05-26 00-33-50 Screenshot from 2020-05-26 00-34-17

tin1254 commented 4 years ago

Another question is how you generate the Graylm with the coordinates from laserscanner? I found the scalings of data from laserscanner vs Graylm are different Screenshot from 2020-05-26 00-41-37

In the screenshot, the scaling of x-axis is clearly different by looking at the pallet

IhabMohamed commented 4 years ago

Hi @tin1254

sorry for the delayed response. Right now, I do not have a Matlab installed in my machine to check. However, as you know, we have 340 scans that have been converted into 340 images, which after that resized into 250x250 pixels. Afterward, these images are rotated by 90 and -90 degrees. Moreover, please, consider the RGB images given here (especially, Pallet250size folder), because I do not remember whether we updated the ROI for Gray images or not. Also, the table of images and their ROI generated by the Image Labeler App might not have the same order as your original data. Here is an example: inside RGBImages.tar.gz /PalletRGBImages/Pallet250size/PalletDataWithRotation_RGBImage_NoOverlaps.mat

Converting the raw data given by the laser scanner from Polar to Cartesian coordinates is given here.

I hope that helps a little bit!

IhabMohamed commented 4 years ago

Hi @tin1254

Could you please close this issue if we answered your question?

Regards,

tin1254 commented 4 years ago

@IhabMohamed Thanks for the answer. However, I still can't find solution to the correspondence between the ROI and the raw laser scanner data, as I would like to use the sensor data. My observation of this issue was based on AllData/PalletImages/PalletGrayImages.zip, there is a file named GrayPalletDataset.mat, which I believe contains the ROI of the gray images. I think the order was just somehow permuted by Image Labeler App

IhabMohamed commented 4 years ago

@tin1254, as I told you, we have used the RGB images. I think their order might be permuted as well. So, if you are interested in the ROIs of the raw data, it might be easier to generate them using Image Labeler App.