Image-Adaptive-3DLUT
Image-Adaptive-3DLUT copied to clipboard
HuiZeng
Learning Image-adaptive 3D Lookup Tables for High Performance Photo Enhancement in Real-time
Thank you for your excellent work. when I used the adaptive 3dlut for video enhancement. I have a question about some smooth areas, like as sky, which has some banding....
RuntimeError: ONNX export failed: Couldn't export python oprator TrilinerpolationFunction
In file included from /content/libcom/libcom/image_harmonization/source/trilinear_cpp/src/trilinear_cuda.cpp:1: /content/libcom/libcom/image_harmonization/source/trilinear_cpp/src/trilinear_kernel.h:4:10: fatal error: THC/THC.h: No such file or directory 4 | #include | ^~~~~~~~~~~ compilation terminated. error: command '/usr/bin/x86_64-linux-gnu-gcc' failed with exit code 1 I...
您好，在运行您的训练代码时，首先报了ModuleNotFoundError: No module named 'trilinear'这个错误，我通过pip安装trilinear后，再次运行报错AttributeError: module 'trilinear' has no attribute 'forward'；当我卸载trilinear后，并且确保sh setup.sh运行成功，再次运行训练代码，报错ModuleNotFoundError: No module named 'trilinear'；请问这种问题应该如何解决呢？是什么原因导致的？ 备注运行环境：cuda10.2 pytorch1.8.1 python3.7
Can you provide codes for converting hdrplus dataset to XYZ colorspace? or tell me how to do it. thx
关于LUT的问题
结合论文、代码，我的理解是LUT0、LUT1、LUT2是可学习的参数，整个模型最终的目的就是优化目标函数获取一个较好的LUT(LUT=w0*LUT0+w1*LUT1+w2*LUT2)；请问可以把LUT0、LUT1、LUT2看作是可学习的参数吗？
训练log： [Epoch 392/400] [Batch 4499/4500] [psnr: 24.850211, tv: 0.001214, wnorm: 1.216050, mn: 0.000000] ETA: 0:07:31.872150 [PSNR: 24 .986909] [max PSNR: 25.268663, epoch: 215] [Epoch 393/400] [Batch 4499/4500] [psnr: 24.900620, tv:...
使用官方的 5 层 CNN，训练结果 使用 ResNet18，训练结果 使用 EfficientNet_b0， 使用 EfficientNet_v2 使用 neighborhood attention network， 请问有哪位遇到过类似的问题吗？
如果我想要更改CNN模型的kernel size，例如改成5*5大小，以便看是否有更高的PSNR。 我跟改了model_x 檔案第52行"layers = [nn.Conv2d(in_filters, out_filters, 5, stride=2, padding=1)]" 跑出了錯誤訊息 "RuntimeError: Calculated padded input size per channel: (7 x 7). Kernel size: (8 x 8). Kernel size can't be...
