search

ding3820 / MIMO-VRN

Video Rescaling Networks with Joint Optimization Strategies for Downscaling and Upscaling (CVPR 2021)
44 stars 4 forks source link

about the lmdb #2

Open pnnngeorge opened 2 years ago

pnnngeorge commented 2 years ago

I stuck with the problem where "buf = txn.get(key.encode('ascii'))", it told me buf is None while both key and txn have vaule, so is it due to the key-value problem, which I mean the value buf can not get from txn using key 'key'. Can you help me? Thank you!

ding3820 commented 2 years ago

Hi, I guess the problem might come from the dataset. If the generated pickle file and lmdb is matched like what I did with Vimeo dataset and with the datascript I provided, it should be fine. I would suggest you run the code without using lmdb if you want to customize your dataset. If you do really want to use lmdb, checkout the latest implementation of BasicSR. They have updated the lmdb implementation whereas the one in this repo is still the deprecated version. You might need to customize your dataloader as well but base on BasicSR should be no problem. Good luck.

pnnngeorge commented 2 years ago

Hi, I guess the problem might come from the dataset. If the generated pickle file and lmdb is matched like what I did with Vimeo dataset and with the datascript I provided, it should be fine. I would suggest you run the code without using lmdb if you want to customize your dataset. If you do really want to use lmdb, checkout the latest implementation of BasicSR. They have updated the lmdb implementation whereas the one in this repo is still the deprecated version. You might need to customize your dataloader as well but base on BasicSR should be no problem. Good luck.

I want to recur MIMO-VRN, so I also use dataset vimeo90k, and I run your code 'create_lmdb.py', then I got the problem. I referred to BasicSR also but there's difference about meta_info, you use '.pkl' while BasicSR use '.txt'. Thanks anyway!

pnnngeorge commented 2 years ago

Hi, I guess the problem might come from the dataset. If the generated pickle file and lmdb is matched like what I did with Vimeo dataset and with the datascript I provided, it should be fine. I would suggest you run the code without using lmdb if you want to customize your dataset. If you do really want to use lmdb, checkout the latest implementation of BasicSR. They have updated the lmdb implementation whereas the one in this repo is still the deprecated version. You might need to customize your dataloader as well but base on BasicSR should be no problem. Good luck.

I want to recur MIMO-VRN, so I also use dataset vimeo90k, and I run your code 'create_lmdb.py', then I got the problem. I referred to BasicSR also but there's difference about meta_info, you use '.pkl' while BasicSR use '.txt'. Thanks anyway!

sorry the problem occurred when I was running train.py, it told me buf is None just like what I said earlier

pnnngeorge commented 2 years ago

Do you have wechat or can I add your wechat to communicate with you about this problem?😂

ding3820 commented 2 years ago

Just pm you by your email