search

Xiangyu-Gao / Radar-multiple-perspective-object-detection

Codes and template data for paper "RAMP-CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition"
MIT License
51 stars 18 forks source link

Raw ADC Radar data UWCR dataset several questions about the data to train on. #16

Open jimvermunt opened 8 months ago

jimvermunt commented 8 months ago

In the paper "RAMP-CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition" it is mentioned that the training and test set have 67,198 and 25,098 training and testing sets, totaling 92,296 frames.

I checked with the raw radar set and only got 23 sequences in total, which each consist of 899 frames, totaling 20,677 frames. This is only 1/5th of the data that is mentioned in the paper. Is the raw radar set a subset of the data where the RAMP-CNN model is trained?

Could you elaborate more on what the differences are?

jimvermunt commented 8 months ago

And what are the sequences used in the following set for training and testing?

sequences_all = ['2019_05_28_pm2s012','2019_05_29_bcms000', '2019_04_09_css1000', '2019_04_30_cm1s000', '2019_05_29_pcms005', '2019_05_09_mlms003', '2019_04_30_mlms000', '2019_05_09_cm1s003', '2019_04_09_pms1000', '2019_04_09_pms3000', '2019_05_09_pbms004', '2019_05_29_mlms006', '2019_05_09_bm1s007', '2019_04_09_pms2000', '2019_04_09_bms1000', '2019_05_09_pcms002', '2019_05_29_pbms007', '2019_04_30_mlms001', '2019_04_30_pbms002', '2019_04_09_cms1000', '2019_04_30_pcms001', '2019_04_30_pbss000', '2019_05_29_cm1s014']
jimvermunt commented 8 months ago

From the raw radar set UWCR in the sequence 2019_04_09_pms2000 it has 0-897 images, 2-899 raw frames, and 3-897 text labels. How does it come that these are misaligned?

This also holds for other sequences but they have different sizes of misalignment.

Dir sequence ../data/Automotive/2019_05_29_bcms000
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
1
Dir sequence ../data/Automotive/2019_04_09_css1000
Number of images: 897
Number of radar frames: 897
Number of text labels: 894
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000896.jpg
Id of first and last radar frame: 000003.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000896.csv
Diff first image and first radar frame: -3
Diff first image and first text label: -3
Diff last image and last radar frame: -3
Diff last image and last text label: 0
2
Dir sequence ../data/Automotive/2019_04_30_cm1s000
Number of images: 900
Number of radar frames: 898
Number of text labels: 896
Diff images and radar frames: 2
Diff images and text labels: 4
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000898.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 1
3
Dir sequence ../data/Automotive/2019_05_29_pcms005
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
4
Dir sequence ../data/Automotive/2019_05_09_mlms003
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
5
Dir sequence ../data/Automotive/2019_04_30_mlms000
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
6
Dir sequence ../data/Automotive/2019_05_09_cm1s003
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
7
Dir sequence ../data/Automotive/2019_04_09_pms1000
Number of images: 898
Number of radar frames: 898
Number of text labels: 895
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000897.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000897.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: -2
Diff last image and last text label: 0
8
Dir sequence ../data/Automotive/2019_04_09_pms3000
Number of images: 898
Number of radar frames: 898
Number of text labels: 895
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000897.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000897.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: -2
Diff last image and last text label: 0
9
Dir sequence ../data/Automotive/2019_05_09_pbms004
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
10
Dir sequence ../data/Automotive/2019_05_29_mlms006
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
Xiangyu-Gao commented 8 months ago

In the paper "RAMP-CNN: A Novel Neural Network for Enhanced Automotive Radar Object Recognition" it is mentioned that the training and test set have 67,198 and 25,098 training and testing sets, totaling 92,296 frames.

I checked with the raw radar set and only got 23 sequences in total, which each consist of 899 frames, totaling 20,677 frames. This is only 1/5th of the data that is mentioned in the paper. Is the raw radar set a subset of the data where the RAMP-CNN model is trained?

Could you elaborate more on what the differences are?

Yes, this raw radar set is a subset of the data where the RAMP-CNN model is trained & tested.

Xiangyu-Gao commented 8 months ago

And what are the sequences used in the following set for training and testing?

sequences_all = ['2019_05_28_pm2s012','2019_05_29_bcms000', '2019_04_09_css1000', '2019_04_30_cm1s000', '2019_05_29_pcms005', '2019_05_09_mlms003', '2019_04_30_mlms000', '2019_05_09_cm1s003', '2019_04_09_pms1000', '2019_04_09_pms3000', '2019_05_09_pbms004', '2019_05_29_mlms006', '2019_05_09_bm1s007', '2019_04_09_pms2000', '2019_04_09_bms1000', '2019_05_09_pcms002', '2019_05_29_pbms007', '2019_04_30_mlms001', '2019_04_30_pbms002', '2019_04_09_cms1000', '2019_04_30_pcms001', '2019_04_30_pbss000', '2019_05_29_cm1s014']

'2019_05_28_pm2s012' is in the testing set for ramp-cnn paper, and others are in the training set.

Xiangyu-Gao commented 8 months ago

From the raw radar set UWCR in the sequence 2019_04_09_pms2000 it has 0-897 images, 2-899 raw frames, and 3-897 text labels. How does it come that these are misaligned?

This also holds for other sequences but they have different sizes of misalignment.

Dir sequence ../data/Automotive/2019_05_29_bcms000
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
1
Dir sequence ../data/Automotive/2019_04_09_css1000
Number of images: 897
Number of radar frames: 897
Number of text labels: 894
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000896.jpg
Id of first and last radar frame: 000003.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000896.csv
Diff first image and first radar frame: -3
Diff first image and first text label: -3
Diff last image and last radar frame: -3
Diff last image and last text label: 0
2
Dir sequence ../data/Automotive/2019_04_30_cm1s000
Number of images: 900
Number of radar frames: 898
Number of text labels: 896
Diff images and radar frames: 2
Diff images and text labels: 4
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000898.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 1
3
Dir sequence ../data/Automotive/2019_05_29_pcms005
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
4
Dir sequence ../data/Automotive/2019_05_09_mlms003
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
5
Dir sequence ../data/Automotive/2019_04_30_mlms000
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
6
Dir sequence ../data/Automotive/2019_05_09_cm1s003
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
7
Dir sequence ../data/Automotive/2019_04_09_pms1000
Number of images: 898
Number of radar frames: 898
Number of text labels: 895
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000897.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000897.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: -2
Diff last image and last text label: 0
8
Dir sequence ../data/Automotive/2019_04_09_pms3000
Number of images: 898
Number of radar frames: 898
Number of text labels: 895
Diff images and radar frames: 0
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000897.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000897.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: -2
Diff last image and last text label: 0
9
Dir sequence ../data/Automotive/2019_05_09_pbms004
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0
10
Dir sequence ../data/Automotive/2019_05_29_mlms006
Number of images: 900
Number of radar frames: 898
Number of text labels: 897
Diff images and radar frames: 2
Diff images and text labels: 3
Id of first and last image: 0000000000.jpg 0000000899.jpg
Id of first and last radar frame: 000002.mat 000899.mat
Id of first and last text label: 0000000003.csv 0000000899.csv
Diff first image and first radar frame: -2
Diff first image and first text label: -3
Diff last image and last radar frame: 0
Diff last image and last text label: 0

Use the filename as the timestamp to match the radar data, image, and labels. You can disregard the redundant ones.