cyy_othermind/fusion实验记录/实验对比.md

实验对比

2026.02.25

训练yaml

'experiment' :

  # 'path_dataset' : '/mnt/data/cdy/project/dataset/FUSION'
  # 'path_result': '/mnt/data/cdy/data2/results/FUSIONLCD'

  # 'path_dataset' : 'E:\work\Project\dataset\FUSION'
  # 'path_result' : 'E:\work\Project\results\FUSIONLCD\bev2'
  
  'path_dataset' : '/home/adlab36/chenyouyuan/FUSIONLCD'
  'path_result': '/home/adlab36/chenyouyuan/FUSIONLCD/result'
  'train_flag' : 1
  'validate_flag' : 0
  'test_flag' : 0
  'flag' : 'fusion' 
  'cuda' : 1
  # TRAINING
  'epochs' : 200
  'batchsize' : 6
  'learning_rate' : 1.e-3
  'beta1' : 0.9
  'beta2' : 0.999
  'eps' : 1.e-8
  'weight_decay' : 5.e-6
  'load_model' : 0
  #FUSION
  # 'last_model' : '/data4/caodanyang/results/FUSIONLCD/08310/models/checkpoint_079.pth.tar'
  #BEV
  # 'last_model' : '/data4/caodanyang/results/FUSIONLCD/bev_09030/models/checkpoint_066.pth.tar'
  #BEV+EP
  'last_model' : '/home/adlab36/chenyouyuan/FUSIONLCD/result/log/models/checkpoint_199.pth.tar'
  #DATASET
  'train' : 0,5,6,7,9
  'validate' : 8,50,54,55,56,59
  'test' : 8,50,54,55,56,59
  'voxel_num' : 15000
  'voxel_max_points' : 100
  'voxel_sample' : 'top'
#  'bev_range' : -51.2,-51.2,-2.5,51.2,51.2,1.5
#  'bev_resolution' : 0.16
#  'bev_range' : -64,-64,-2.5,64,64,1.5
#  'bev_resolution' : 0.2
  'bev_range' : -32,-32,-2.5,32,32,1.5
  'bev_resolution' : 0.2

  # NETWORK PARAMS
  'kpts_number_bev' : 150
  'kpts_number_img' : 0
  'cluster_num_bev' : 16
  'cluster_num_img' : 0
  'cluster_num_fusion' : 16
  'sinkhorn_iter' : 5
  'vlad_size' : 256
  # LOSS
  'loop_file' :  'loop_GT_4m'
  'trip_margin' : 0.5
  'negetative_selsector' : 'random'

测验yaml

'experiment' :

  # 'path_dataset' : '/mnt/data/cdy/project/dataset/FUSION'
  # 'path_result': '/mnt/data/cdy/data2/results/FUSIONLCD'

  # 'path_dataset' : 'E:\work\Project\dataset\FUSION'
  # 'path_result' : 'E:\work\Project\results\FUSIONLCD\bev2'
  
  'path_dataset' : '/home/adlab36/chenyouyuan/FUSIONLCD'
  'path_result': '/home/adlab36/chenyouyuan/FUSIONLCD/result'
  'train_flag' : 0
  'validate_flag' : 1
  'test_flag' : 1
  'flag' : 'fusion' 
  'cuda' : 1
  # TRAINING
  'epochs' : 200
  'batchsize' : 6
  'learning_rate' : 1.e-3
  'beta1' : 0.9
  'beta2' : 0.999
  'eps' : 1.e-8
  'weight_decay' : 5.e-6
  'load_model' : 1
  #FUSION
  # 'last_model' : '/data4/caodanyang/results/FUSIONLCD/08310/models/checkpoint_079.pth.tar'
  #BEV
  # 'last_model' : '/data4/caodanyang/results/FUSIONLCD/bev_09030/models/checkpoint_066.pth.tar'
  #BEV+EP
  'last_model' : '/home/adlab36/chenyouyuan/FUSIONLCD/result/log/models/checkpoint_199.pth.tar'
  #DATASET
  'train' : 0,5,6,7,9
  'validate' : 8,50,54,55,56,59
  'test' : 8,50,54,55,56,59
  'voxel_num' : 15000
  'voxel_max_points' : 100
  'voxel_sample' : 'top'
#  'bev_range' : -51.2,-51.2,-2.5,51.2,51.2,1.5
#  'bev_resolution' : 0.16
#  'bev_range' : -64,-64,-2.5,64,64,1.5
#  'bev_resolution' : 0.2
  'bev_range' : -32,-32,-2.5,32,32,1.5
  'bev_resolution' : 0.2

  # NETWORK PARAMS
  'kpts_number_bev' : 150
  'kpts_number_img' : 0
  'cluster_num_bev' : 16
  'cluster_num_img' : 0
  'cluster_num_fusion' : 16
  'sinkhorn_iter' : 5
  'vlad_size' : 256
  # LOSS
  'loop_file' :  'loop_GT_4m'
  'trip_margin' : 0.5
  'negetative_selsector' : 'random'

测验结果


与论文对比

同级目录下表格文件