irfanICMLL / structure_knowledge_distillation

The official code for the paper 'Structured Knowledge Distillation for Semantic Segmentation'. (CVPR 2019 ORAL) and extension to other tasks.

irfanICMLL/structure_knowledge_distillation Issues

R.I.P.
Updated 5 months ago8
How can I run build.py in pytorch1.9?
Closed 3 years ago10
I can't get your mIoU on cityscapes dataset.
Closed 4 years ago3
softmax should be calculated along the last dimension(-1), not dim 1
Closed 9 months ago1
Training from scratch reached the performance similar with that when initializing from ImageNet
Updated a year ago1
training log
Updated a year ago1
How to run on my own dataset?
Updated 2 years ago4
How to apply KD for two input images?
Updated 2 years ago
__ext.so error: undefined symbol: __cudaRegisterFatBinaryEnd
Updated 2 years ago1
Example training method for Depth Estimation VNL network
Closed 3 years ago2
training fails with "RuntimeError: cuda runtime error (11) : invalid argument at THCGeneral.cpp:405"
Updated 3 years ago2
/data/ckj/SKD-master/libs/_ext/__ext.so: failed to map segment from shared object
Updated 3 years ago
code error
Updated 3 years ago
concat RGB images and seg maps as input of discriminator
Updated 3 years ago
Holistic loss
Updated 3 years ago2
FileNotFoundError: [Errno 2] No such file or directory: './ckpt/CS_scenes_39326_0.75.pth'
Updated 3 years ago
Pretrained student?
Closed 3 years ago2
build error
Closed 4 years ago2
Why the patch size turn out to be half of the original feature map size in CriterionPairWiseforWholeFeatAfterPool?
Closed 3 years ago3
Can you provide the ResNet(1.0) baseline train log?
Closed 3 years ago1
Train script
Closed 3 years ago1
Hello, I also cannot reproduce the results
Closed 3 years ago6
About the input of discriminator
Updated 3 years ago
Pixel-wise loss
Closed 4 years ago5
How to apply pair-wise distillation on depth prediction task
Updated 4 years ago1
the pixelwise loss never decreased during training
Updated 4 years ago4
about the code
Updated 4 years ago2
Puzzled about this loop
Closed 4 years ago1
D_loss becomes NAN during training
Closed 4 years ago2
codes for PAMI version
Closed 4 years ago2
checkpoint
Closed 4 years ago
Why structure knowledge distillation is effective?
Updated 4 years ago2
TypeError: add_() received an invalid combination of arguments - got (complex, Tensor)
Closed 4 years ago5
ESPNet training using 'master' & 'cvpr2019' branch
Updated 4 years ago1
The paper use average pooling to aggregate β × C features in one node to be 1 × C, but the code shows max pooling.
Updated 4 years ago1
ESPNet training
Closed 4 years ago2
no models for detection and depth estimation
Closed 4 years ago3
retrain teacher net
Closed 4 years ago2
Build error
Updated 4 years ago1
CUDA Error encountered in <function CompiledLib.bn_forward_cuda at 0x7fef222d0400>
Closed 4 years ago1
Hi, I cannot reproduce the result.
Closed 5 years ago4
Could you provide your training log?
Closed 4 years ago3
Question about the pair-wise loss
Closed 4 years ago1
The source code of FCOS
Closed 4 years ago1
pixel-wise loss dominates generator training
Closed 4 years ago3
about pair-wise and pixel-wise distillation for FCOS
Closed 4 years ago4
run_train_val.sh: --pi: not found
Closed 4 years ago2
nan loss value
Closed 5 years ago4
TypeError: add_() received an invalid combination of arguments
Closed 5 years ago2
Where is Discriminator weights
Closed 5 years ago1