msracver/Deformable-ConvNets

Issues

• About the range of offset

  #302 opened a year ago by Chris-city
  1

• Is it useful to use dcn on 1x1conv?

  #299 opened 2 years ago by kirkzZ
  1

• PyTorch version of deformable convolution does not exist

  #301 opened a year ago by coloedrainbow
  1

• Question about num_deformable_group

  #268 opened 5 years ago by chongruo
  3

• src/cpu/dcn_v2_cpu.cpp:227:55: 错误：‘THFloatBlas_gemv’在此作用域中尚未声明

  #300 opened 2 years ago by oszn
  0

• convert to onnx

  #298 opened 2 years ago by Eyesfront-ZJG
  0

• How to get the receptive field?

  #282 opened 3 years ago by kirkzZ
  4

• How to get the sampling points?

  #283 opened 3 years ago by kirkzZ
  4

• Confusion about why a single conv layer is able to learn the offsets

  #252 opened 5 years ago by eezywu
  3

• Why is the magnitude of mask multiplied by 2?

  #280 opened 3 years ago by developer0hye
  1

• Does anyone sovle the problem of over large offset?

  #277 opened 4 years ago by hmjbuaa
  1

• Some questions of the testting process in classification tasks

  #279 opened 4 years ago by crocodilegogogo
  0

• occasions where sampling locations are outside of the image boundary

  #278 opened 4 years ago by vlee-harmonicinc
  0

• DCN for 3D

  #275 opened 4 years ago by doramran
  0

• visualize the deformable convolution of rcnn_dcn

  #276 opened 4 years ago by xuwwwux
  0

• where to embed deform_conv layer?

  #237 opened 5 years ago by satchmo233
  2

• compile DCN v2 fatal error: ../linalg.h: No such file or directory

  #272 opened 5 years ago by jiashu-zhu
  4

• ##

  #261 opened 5 years ago by evorta
  0

• #

  #260 opened 5 years ago by evorta
  0

• The channels' order of offsets layer

  #274 opened 5 years ago by whulc
  0

• error when run deeplab(cityscapes) ValueError: zero-size array to reduction operation minimum which has no identity`

  #273 opened 5 years ago by zyxu1996
  0

• Deeplab Training Detail

  #271 opened 5 years ago by donhk1011
  0

• Has anyone successfully trained Faster-RCNN with Modulated Deformable Conv?

  #270 opened 5 years ago by chongruo
  0

• Can I only use CUDA8.0?

  #269 opened 5 years ago by WHY95666
  0

• mxnet.base.MXNetError: Error in operator fpn_p4_sum:

  #266 opened 5 years ago by herrickli
  1

• undefined symbol: _ZN2at7getTypeERKNS_6TensorE

  #264 opened 5 years ago by xiaoliqiangxxqzz
  2

• dcn = DCN(64, 64, kernel_size=(3,3), stride=1, padding=1, deformable_groups=2).cuda()

  #265 opened 5 years ago by xiaoliqiangxxqzz
  0

• Could you provide results file in json form from DCN v2?

  #262 opened 5 years ago by ArchNew
  0

• Which version of tensorflow should be used?

  #257 opened 5 years ago by wander1985
  0

• raise MXNetError(py_str(_LIB.MXGetLastError()))

  #259 opened 5 years ago by wander1985
  0

• keras deformable

  #258 opened 5 years ago by ICDI0906
  0

• Exception while resuming training: assert isinstance(step, list) and len(step) >= 1 AssertionError

  #248 opened 5 years ago by pervaizniazi
  1

• What does `mAP@0.75` mean?

  #256 opened 5 years ago by jiec-msft
  1

• Is the evaluation correct for pascal segmentation ?

  #255 opened 5 years ago by pengwangucla
  0

• Symbol file to train DCNv2

  #254 opened 5 years ago by miraj7789
  0

• windows10 when i build cython ,someone can help me ?i really don't know how to deal.

  #253 opened 5 years ago by gdut3118007086
  0

• why use conv-layer to predict conv offset while fc-layer predict pool offset?

  #250 opened 5 years ago by hust-kevin
  1

• why not use im2col_step in ModulatedDeformableConvolutionOp?

  #251 opened 5 years ago by Google1234
  0

• Question about dcnv2 ops

  #249 opened 5 years ago by henrylee2570
  0

• RCNNLogLoss not decreas

  #247 opened 5 years ago by zhengye1995
  0

• RCNNLogLoss not decreasing while traing

  #238 opened 5 years ago by yohohohoho
  4

• Could you explain why mask generated by Sigmoid in mdcn is doubled?

  #246 opened 5 years ago by JrPeng
  1

• this gemm operation seemes useless

  #245 opened 5 years ago by BIGKnight
  1

• RpnL1loss is very high when I use 8gpus and change lr to 0.01

  #244 opened 5 years ago by anorthman
  0

• loss

  #243 opened 5 years ago by anorthman
  0

• In readme, for deeplab, we use 4 GPUs for all experiments

  #242 opened 5 years ago by KeyKy
  0

• Seems the implementation of offsets is different from the original paper?

  #241 opened 5 years ago by SUZhaoyu
  1

• A bug？

  #236 opened 5 years ago by Google1234
  1

• why slicing the input data at the batch dimension ?

  #240 opened 5 years ago by Google1234
  0

• No module named skimage.draw

  #239 opened 5 years ago by yfnn
  1