sunwei925 / simplevqa Goto Github PK

官方的表格填了没有回应，请问可以分享下么，非常感谢。

File "/data4/xxx/base37/lib/python3.7/site-packages/torch/nn/parallel/parallel_apply.py", line 64, in _worker
output = module(*input, **kwargs)
File "/data4/xxx/base37/lib/python3.7/site-packages/torch/nn/modules/module.py", line 1190, in _call_impl
return forward_call(*input, **kwargs)
File "/data4/xxx/VideoQualiry/SimpleVQA/model/UGC_BVQA_model.py", line 263, in forward
return self._forward_impl(x, x_3D_features)
File "/data4/xxx/VideoQualiry/SimpleVQA/model/UGC_BVQA_model.py", line 252, in _forward_impl
x = torch.cat((x, x_3D_features), dim = 1)
RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 7 but got size 8 for tensor number 1 in the list.

您好 我想在KoNViD-1k数据集上训练模型，但是train_baseline.py文件没有在仓库中找到，然后train.py文件是针对LSVQ数据集训练的，请问方便提供一下train_baseline.py文件吗

May I ask if the author can provide the video sets in LSVQ CSV or download link? In the LSVQ link in ReadME, there don't seem to be so many LSVQ video sets downloaded. Thank you for your reply!

Hello, I want to train your model on the KonViD-1k database, I used your script extract_frame_konvid1k.py, however, I think the KoNViD-1k_data.mat file contains different names of the dataset, ('8171831850_cut_centercrop_960x540_8s.mp4').

Hello,where is train_baseline.py?

您好，想复现一下您在KoNViD-1k数据库上的结果，没有找到数据集划分的csv文件，是否可以更新一下仓库，万分感谢！！！

你好，在introduction部分，“Instead of using the pretrained model to extract the spatial features in the previous studies, we propose to train an end-to-end spatial feature extraction network...”,说的是没有使用pretrained的模型。在implement details部分说的是'The weights of the ResNet50 are initialized by training on the ImageNet dataset'.请问这个是怎么理解的？ 谢谢大佬