hustvl/SparseInst

Issues

• Could you tell me where can I find the code for multiplying the DICE loss with the classification score that you mentioned in your paper?

  #128 opened 3 months ago by Run-19
  0

• Is lightweight backbone like mobilenet available in the future?

  #121 opened 2 years ago by wooAo
  1

• Bug of `demo.py`

  #100 opened 2 years ago by leftthomas
  3

• loss_mask:nan

  #126 opened a year ago by Zhounjau
  0

• training details

  #124 opened a year ago by lsm140
  2

• When running train_net.py, an error occurs.

  #125 opened a year ago by kirorotarzan
  0

• 在COCO数据集上使用sparse_inst_r50vd_dcn_giam_aug.yaml重头训练，精度并没有达到37.9，低了十几个点

  #108 opened 2 years ago by 116022017144
  2

• num_masks

  #99 opened 2 years ago by 18219716332
  2

• I think the performance of your model is not as high as Condinst

  #123 opened a year ago by Originlightwkp
  0

• 关于样本分配的顺序问题（或者说训练为什么可以收敛？）

  #113 opened 2 years ago by feiyuhuahuo
  1

• 关于可视化IAM的问题

  #111 opened 2 years ago by GTrui6
  5

• Single kernel crossponding to multiple object

  #122 opened 2 years ago by Lincoln12w
  0

• RuntimeWarning: invalid value encountered in intersection

  #120 opened 2 years ago by zyqdragon
  0

• hyper-parameters not same with paper

  #118 opened 2 years ago by leftthomas
  0

• Why init weights like that?

  #119 opened 2 years ago by leftthomas
  0

• I can't find the video result

  #83 opened 2 years ago by BibibNanana
  12

• train on cocodataset ( AP is so low )

  #117 opened 2 years ago by Wooho-Moon
  2

• different results between ONNX model and pt model

  #101 opened 2 years ago by zoey9628
  2

• error in training by using voc2012

  #116 opened 2 years ago by 116022017144
  0

• About experiments on Cityscapes

  #115 opened 2 years ago by cici-999
  0

• Please help to get started with COCO-style custom datasets and classes

  #114 opened 2 years ago by thusinh1969
  0

• APs is nan

  #112 opened 2 years ago by 18219716332
  0

• images without any annotation

  #97 opened 2 years ago by xjsxujingsong
  4

• 这个文件/root/anaconda3/envs/sparse/lib/python3.7/site-packages/fvcore/transforms/transform.py对polygon数据的处理是怎么样的

  #110 opened 2 years ago by 116022017144
  0

• 在COCOTrain2017数据集训练，Val2017测试的AP是27.00，为什么这么低呢?

  #103 opened 2 years ago by 1032697379
  4

• [Checkpointer] Loading from pretrained_models/resnet50d_ra2-464e36ba.pth fvcore.common.checkpoint WARNING: Some model parameters or buffers are not found in the checkpoint:

  #109 opened 2 years ago by 116022017144
  0

• train the model not use fp16

  #107 opened 2 years ago by 116022017144
  0

• onnx model output dimension.

  #106 opened 2 years ago by YaoJiawei329
  0

• ValueError: matrix contains invalid numeric entries for binary classification

  #84 opened 2 years ago by kirillkoncha
  3

• Output Resolution

  #102 opened 2 years ago by marcown
  1

• demo.py error

  #105 opened 2 years ago by Xuying2000
  1

• demo.py error

  #104 opened 2 years ago by Xuying2000
  0

• ( FIXED )WHY FPS is so low on my datasets ??

  #98 opened 2 years ago by g110056012
  2

• SpareInst based on mmdet

  #95 opened 2 years ago by zaneoo
  3

• Why does 'thing_classes' of custom COCO format dataset always wrong?

  #96 opened 2 years ago by China-EvilDevil
  0

• The problem of two gpus training

  #93 opened 2 years ago by cici-999
  2

• Quick tutorial for newbs?

  #94 opened 2 years ago by nathanieltagg
  2

• can release more model like cspdarknet-sparseinst

  #91 opened 2 years ago by DDGRCF
  4

• Help wanted for quantized int8 model

  #88 opened 2 years ago by Stephenzza
  4

• RuntimeError: Given input size: (256x9x27). Calculated output size: (256x0x2). Output size is too small

  #92 opened 2 years ago by YaoJiawei329
  7

• Setting thresholds during inference

  #86 opened 2 years ago by neo031995
  4

• Why AP so low?

  #87 opened 2 years ago by BibibNanana
  1

• xperiment with cityscapes dataset

  #90 opened 2 years ago by cici-999
  0

• Custom dataset training: Change paths

  #79 opened 2 years ago by kirillkoncha
  13

• CSPDarknet53 as a backbone

  #82 opened 2 years ago by koolvn
  5

• How to make the output image into a binary image？

  #85 opened 2 years ago by BibibNanana
  3

• No penalty for unmatched prediction

  #77 opened 2 years ago by zoucheng1991
  2

• mAP equals 0 after 10 epochs of training

  #81 opened 2 years ago by RocketFlash
  2

• how to improve the inference speed?

  #78 opened 2 years ago by SiChuanJay
  3

• This issue is mistakenly created. Please remove.

  #80 opened 2 years ago by ysnb-taniguchi
  1