NVlabs/SegFormer

Issues

• how to config mmcv-full-1.2.7 on Windows

  #146 opened 3 months ago by honghongym
  2

• raise StopIteration

  #147 opened 3 months ago by dawang-11
  0

• Potential mistake in SegFormer model: `patch_size` argument in SegFormer model not being used.

  #141 opened 6 months ago by jonasdieker
  1

• Implementation issues of the Efficient Self-Attention module

  #148 opened a month ago by K1t3
  0

• SegFormerB1 CityScapes - CUDA error: an illegal memory access was encountered

  #126 opened a year ago by adriengoleb
  1

• What does "type = IMTRv21_5" mean?

  #107 opened 2 years ago by Buling-Knight
  3

• [Reward] How to freeze and finetune?

  #145 opened 5 months ago by Meoooww
  0

• single-gpu training is ok, but Multi-gpu training has some wrong result

  #144 opened 5 months ago by zhedahe
  0

• About the ImageNet1k pretrain

  #102 opened 2 years ago by zhanggj821
  5

• Pre-training info

  #116 opened a year ago by nargenziano
  5

• About the PatchEmbedding

  #98 opened 2 years ago by SimonServant
  2

• Incorrect Visualization

  #117 opened a year ago by Jayden9912
  2

• Use of different optimizers in SegFormer

  #142 opened 6 months ago by anirbala98
  0

• RuntimeError: Only tuples, lists and Variables supported as JIT inputs/outputs. Dictionaries and strings are also accepted but their usage is not recommended. But got unsupported type numpy.ndarray

  #140 opened 6 months ago by hengyanchen
  0

• Model file not found

  #139 opened 7 months ago by daviscsuft
  0

• Permission Issues with Google Drive Link for MiT ImageNet Weights

  #136 opened 7 months ago by lhoyer
  1

• Geforce RTX 3060 must use torch11.0+(Windows)

  #104 opened 2 years ago by Buling-Knight
  4

• Could you share pretrain recipe?

  #138 opened 7 months ago by sipie800
  0

• Where can I set parameters such as epoch?

  #135 opened 8 months ago by Meoooww
  1

• How to save the best checkpoint file?

  #137 opened 7 months ago by Meoooww
  0

• Questions about the Effective Receptive Field

  #134 opened 8 months ago by lcy8417
  0

• Issue about setting number of GPUs for evaluation during training process

  #133 opened 9 months ago by Yonghao-Tan
  0

• k v compression in attention may causes small targets and detail to be lost？

  #132 opened 9 months ago by swjtulinxi
  0

• keyError 'data_time'

  #131 opened 10 months ago by JJY0710
  1

• License of SegFormer pretrained weights available on other platforms

  #130 opened a year ago by alzaia
  0

• ImportError: cannot import name 'print_log' from 'mmcv'

  #129 opened a year ago by alon2U
  0

• How to convert the model to tensorrt or onnx?

  #114 opened a year ago by yuchenlichuck
  2

• How can Fig. 6 be reproduced?? (Effective Receptive Field)

  #123 opened a year ago by Loparch
  1

• The result of Segformer-b5 on Cityscapes

  #128 opened a year ago by cht619
  0

• What is the image size used to calculate the flops in Table 1?

  #127 opened a year ago by wzhiyan
  0

• ADE B3 model SS mIoU doesn't match paper Table 1.

  #125 opened a year ago by gathierry
  3

• TypeError: `model` must be a `LightningModule` or `torch._dynamo.OptimizedModule`, got `SegformerForSemanticSegmentation`ON CLOAB

  #124 opened a year ago by Wang-taoshuo
  0

• ModuleNotFoundError: No module named 'mmcv._ext'

  #122 opened a year ago by bekhzod-olimov
  2

• why does the model have to pre-trained on ImageNet-1k datasets? Did you test results directly on ADE20K without using pre-training?

  #119 opened a year ago by 754467737
  1

• Welcome update to OpenMMLab 2.0

  #121 opened a year ago by vansin
  0

• how to calculate models fps in paper

  #118 opened a year ago by Shinsungjun
  1

• Clarifying the Pretraining Procedure

  #120 opened a year ago by gauenk
  1

• Excuse me, how to evaluate the accuracy of Cityscapes dataset？

  #113 opened a year ago by ZzzTD
  0

• Weights for the SegFormer-B5 on COCOStuff

  #112 opened a year ago by edwardyehuang
  0

• Dataset Creation

  #111 opened 2 years ago by FatemaD1
  0

• torch onnx export

  #96 opened 2 years ago by germanjke
  1

• backbone about SegNeXt

  #110 opened 2 years ago by kamen007
  0

• Why does the image named val_mIoU so strange?

  #108 opened 2 years ago by Buling-Knight
  1

• Problems with Docker on remote development server

  #106 opened 2 years ago by mak7ostapenko
  0

• Ade20k dataset val/test problem

  #105 opened 2 years ago by gavin-gqzhang
  1

• Math wrong in Overlap Patch Merging

  #103 opened 2 years ago by timdirr
  0

• The checkpoint can not be used

  #101 opened 2 years ago by 25884
  0

• How the final output is converted to full resolution mask?

  #100 opened 2 years ago by working12
  0

• imagenet train code

  #97 opened 2 years ago by endaoguansanlu
  1

• xxx is not a checkpoint file

  #99 opened 2 years ago by toohappygirl
  0