duzzzs / monodlex Goto Github PK

为什么的运行官方的monodle：
(monodle) G:\WWProject\monodle\experiments\example>python ../../tools/train_val.py --config kitti_example.yaml
2022-10-11 10:46:15,326 INFO ################### Training ##################
2022-10-11 10:46:15,326 INFO Batch Size: 16
2022-10-11 10:46:15,326 INFO Learning Rate: 0.001250
然后居然需要过半个小时才出现的训练的epoch进度条，我将原monodle改为单卡运行了，我的是单卡3090
epochs: 0%| | 0/140 [00:00<?, ?it/s]
iters: 0%| | 0/232 [00:00<?, ?it/s]

@DuZzzs
Hi，我这边训完DDP分支下的代码，似乎mAP只有11点多，这个比原来的13.72低了很多，不知道大佬在实现的时候有遇到过这个问题吗？
另外，增加monoflex的edge-fusion模块之后，val上面的精度大概是多少呢？

您好！我看了这两个问题，但还是没搞明白data结构该如何放置数据集。
我的目录如下：
#ROOT
|data/
----|KITTI/
------|training/
--------|ImageSets/
--------|calib/
--------|image_2/
--------|label/
------|testing/
--------|calib/
--------|image_2/

我修改了yaml文件中的root_dir位置为为\monodleX-main\data\KITTI，出现训练第一个echo时，还没开始训练时会报错如下：

File "../../tools/train_val.py", line 88, in
main()
File "../../tools/train_val.py", line 76, in main
trainer.train()
File "F:\xiangmu\monodleX-main\lib\helpers\trainer_helper.py", line 73, in train
self.train_one_epoch()
File "F:\xiangmu\monodleX-main\lib\helpers\trainer_helper.py", line 98, in train_one_epoch
for batch_idx, (inputs, targets, _) in enumerate(self.train_loader):
File "D:\Anaconda3\envs\monodle\lib\site-packages\torch\utils\data\dataloader.py", line 582, in next
return self._process_next_batch(batch)
File "D:\Anaconda3\envs\monodle\lib\site-packages\torch\utils\data\dataloader.py", line 608, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
AssertionError: Traceback (most recent call last):
File "D:\Anaconda3\envs\monodle\lib\site-packages\torch\utils\data_utils\worker.py", line 99, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "D:\Anaconda3\envs\monodle\lib\site-packages\torch\utils\data_utils\worker.py", line 99, in
samples = collate_fn([dataset[i] for i in batch_indices])
File "F:\xiangmu\monodleX-main\lib\datasets\kitti\kitti_dataset.py", line 157, in getitem
objects = self.get_label(index)
File "F:\xiangmu\monodleX-main\lib\datasets\kitti\kitti_dataset.py", line 86, in get_label
assert os.path.exists(label_file)
AssertionError

想请问大佬这还是我数据集放的位置不对的问题吗？看样子好像是无法正确读取标签，还是想问一下大佬data的结构树是什么样的

请问谁有resnet18 backbone的预训练模型吗

作者你好，我用官方的monodle提供的模型和自己重新用kitti数据集训练的模型在kitti city数据集上测试效果很好，但是在自己采集的数据集上测试效果就很差，请问下这个是什么原因

我用的是1920 1080的图片做的测试，感觉中心点坐标偏移的很严重。

Car [email protected], 0.70, 0.70:
bbox AP:98.5218, 92.2878, 82.8832
bev AP:30.7260, 22.9492, 19.8986
3d AP:22.9616, 17.5276, 14.8316
aos AP:97.73, 91.04, 80.95

你好，发现seg loss很难收敛，其中使用单张图片过拟合训练seg loss最后在3以上。

Hello, when reading the label data, if I change the filter conditions of the instance level, does it mean that the distance of the model I trained focus is different, and will the level filtering hyperparameter affect the evaluation process?

您好，我用最新的 main分支下的代码单卡 batch_size=16训练 ，训练的结果离readme离好像有比较大的差距

random_seed: 444
log_dir: 'work_dirs/test_xyn/'

dataset:
  type: &dataset_type 'KITTI'
  batch_size: 16
  use_3d_center: True
  class_merging: False
  use_dontcare: False
  bbox2d_type: 'anno'   # 'proj' or 'anno'
  meanshape: False      # use predefined anchor or not
  writelist: ['Car']
  random_flip: 0.5
  random_crop: 0.5
  scale: 0.4
  shift: 0.1

model:
  type: 'centernet3d'
  backbone: 'dla34'
  neck: 'DLAUp'
  num_class: 3

optimizer:
  type: 'adamw'
  lr: 0.0002
  weight_decay: 0.00001

lr_scheduler:
  warmup: True  # 5 epoches, cosine warmup, init_lir=0.00001 in default
  decay_rate: 0.1
  decay_list: [90, 120]

trainer:
  max_epoch: 140
  gpu_ids: 0
  save_frequency: 10 # checkpoint save interval (in epoch)
  # resume_model: 'checkpoints/checkpoint_epoch_70.pth'


tester:
  type: *dataset_type
  mode: single   # 'single' or 'all'
  checkpoint: 'checkpoints/checkpoint_epoch_140.pth'  # for 'single' mode
  checkpoints_dir: 'checkpoints'  # for 'all' model
  threshold: 0.2  # confidence filter

请问是将head中depthHw2, depth[;,;,0]为深度值，depth[:,:,1]这个直接就是为深度方差，直接用作拉普拉斯不确定度的方差