Pang-Yatian/Point-MAE

Issues

• Compile Error

  #54 opened 3 months ago by Research-Code-123
  2

• About sample strategy during pretrain

  #55 opened 3 months ago by JiaweiMao12135
  1

• Running Error

  #53 opened 3 months ago by Research-Code-123
  0

• About pos_emd and attention distribution

  #52 opened 4 months ago by JiaweiMao12135
  1

• Finetuning modelnet with 4090 GPU card, DataLoader worker (pid 899873) is killed by signal: Bus error.

  #51 opened 4 months ago by Research-Code-123
  1

• Possible extension to outdoor scene-level scenario

  #49 opened 5 months ago by Pamphlett
  1

• Visualization, only partial point clouds of each class are displayed

  #50 opened 4 months ago by SunshineDHW
  1

• point cloud registration

  #48 opened 5 months ago by Dxinz
  1

• Install emd extension with torch>1.11.0

  #42 opened a year ago by lqxisok
  1

• chamfer_dist compile for sm_90 GPU architecture

  #47 opened 7 months ago by saimunur
  1

• have problem when run 'python setup.py install --user'

  #30 opened 8 months ago by ShuhangGe
  2

• Demo code?

  #46 opened 9 months ago by Abbsalehi
  1

• Intuition behind the share-weighted learnable mask token

  #44 opened 9 months ago by nielsRocholl
  1

• ModelNet(8k) accuracy of 1024 PCs

  #45 opened 10 months ago by Harr7y
  2

• (Almost) Identical predicted masked point patches

  #43 opened a year ago by nielsRocholl
  4

• Model architecture

  #41 opened a year ago by bhatipriyanka
  1

• we can't reproduce the results of your experiment,could you give us some suggestion or help,thanks!!!

  #40 opened a year ago by zoesoon
  1

• About finetuning performance

  #39 opened a year ago by sugar-fly
  1

• How to distribute data parallel training the code

  #38 opened a year ago by Chopper-233
  1

• Request for Pretraining Log

  #37 opened a year ago by whalesing
  0

• About the scale factor

  #36 opened a year ago by martianxiu
  1

• Configuration files for finetuning

  #26 opened 2 years ago by ivansipiran
  2

• How many GPUs are needed for pre-training?

  #19 opened 3 years ago by jianghaojun
  4

• Patches

  #35 opened 2 years ago by arndz
  1

• Details on training environment

  #34 opened 2 years ago by meyerjo
  1

• position embedding only operate in visiable patch

  #33 opened 2 years ago by name555difficult
  2

• which ckpts to load for fine-tuning tasks

  #31 opened 2 years ago by kengkeng1
  1

• where to download ShapeNetPart dataset?

  #32 opened 2 years ago by MillyMaMa
  0

• How to train from stratch and evaluate on modelnet40 for object classification?

  #23 opened 2 years ago by TinaJin0228
  6

• The command of finetune on modelnet

  #29 opened 2 years ago by whuhxb
  1

• The result cannot be reproduced on ScanObjectNN_hardest, ScanObjectNN_objectbg, and ShapeNetPart.

  #27 opened 2 years ago by Daybreak-Zheng
  3

• How do I get the results of a few shots?

  #28 opened 2 years ago by Daybreak-Zheng
  0

• not able to import "chamfer"

  #17 opened 2 years ago by krmayankb
  8

• not able to install

  #25 opened 2 years ago by josenachorr
  1

• How to change embedding module output size?

  #24 opened 2 years ago by DanieleMenchetti
  2

• How to open TFBoard file

  #22 opened 2 years ago by WenjieZhao1
  1

• A question about "Few-Shot".

  #14 opened 3 years ago by QWTforGithub
  5

• Questions about Transformer achitecture ?

  #21 opened 2 years ago by zhulf0804
  2

• cls_token during masked reconstruction pretraining

  #20 opened 3 years ago by etaoxing
  1

• How to test on ScanObjectNN?

  #16 opened 3 years ago by Phoebe-ovo
  1

• What is the difference between trans_dim and encoder_dims?

  #15 opened 3 years ago by Phoebe-ovo
  1

• Confusion about the masking strategy and prediction target

  #11 opened 3 years ago by auniquesun
  3

• A question about ShapeNet "train.txt" and "test.txt"

  #13 opened 3 years ago by QWTforGithub
  3

• What do OBJ-BG and OBJ-ONLY mean?

  #12 opened 3 years ago by charlesCXK
  5

• About Visualization

  #10 opened 3 years ago by jlqzzz
  3

• What is the difference between point cloud completion task and point cloud masked reconstruction?

  #9 opened 3 years ago by jlqzzz
  2

• Bugs in modelnet40_8k pretrain model

  #7 opened 3 years ago by cuge1995
  6

• Pre-training Log

  #8 opened 3 years ago by Hiusam
  2

• Reimplement the finetune_scan_objonly.yaml, only 86.78 acc, why?

  #6 opened 3 years ago by cuge1995
  3

• Can't reproduce the experiment on model40 1k

  #5 opened 3 years ago by yanghu819
  31