car detection network. TRained in openpcdet frame work, using PVRECNN model and Innoviz point cloud. We used point cloud which were acquired by Lidar sensor: InnovizOne. InnovizOne is resilient to sunlight and weather conditions, and is able to deliver a rich 3D point cloud at distances up to 250 meters. The maximum FOV is 115〖115〗^0 x〖25〗^0 with frame rate of 10-15 FPS.
GPU resources: 2 NVIDIA GeForce RTX 2080 Ti graphics cards Operation System: Ubuntu 20.04.6 LTS
253 point clouds were used for training and test. 51 point clouds were used validation
Dataset: Folder Name #Frames #pcd files Label counter 2023-03-26-18-53-51_static 4 4 48 2023-03-26-19-28-24_B_MovingSingleCarInParking 100 100 1269 2023-03-26-19-46-19_A 54 54 439 2023-03-26-19-46-19_end 34 34 109 2023-03-26-19-49-25_A 39 39 121 2023-03-26-19-49-25_C 22 22 224 2023-03-26-19-52-04_A 54 54 476
Resuls: 2023-11-20 18:25:55,703 INFO *************** Performance of EPOCH 80 ***************** 2023-11-20 18:25:55,703 INFO Generate label finished(sec_per_example: 0.3311 second). 2023-11-20 18:25:55,704 INFO recall_roi_0.3: 0.991111 2023-11-20 18:25:55,704 INFO recall_rcnn_0.3: 0.993333 2023-11-20 18:25:55,704 INFO recall_roi_0.5: 0.964444 2023-11-20 18:25:55,704 INFO recall_rcnn_0.5: 0.962222 2023-11-20 18:25:55,704 INFO recall_roi_0.7: 0.813333 2023-11-20 18:25:55,704 INFO recall_rcnn_0.7: 0.900000 2023-11-20 18:25:55,704 INFO Average predicted number of objects(51 samples): 8.647