storife / ranet Goto Github PK

Any other location(not Baidu) form which i can get the model?

Thanks

shuoul i down load from davis2016 or davis2017? may you offer me the url?

According to the Figure 2 in the paper, the decoder takes two things in, one is merged similarity maps and mask from 'Merge' module, the other is the features of the current frame extracted directly by Siamese Encoder.

But I can't find anywhere how the decoder uses those two things. Are they concatenated together before going into the Decoder, or there is some sort of operation? Such as convolution, element-wise multiplication or just simple addition?

In your ablation study, you performed the w/o ranking experiment to show the importance of the RAM, but what's the difference between two of them, thanks!

Thanks for your awesome work! BTW, would you plan to release the training code or not?

My GPU is RTX2080ti with 12G, but it is not still satisfied with your code. So what kind of your GPU is?

Hi, I followed the instructions and am trying to generate the predictions for DAVIS17testdev using the RANet_video_multi.pth. However, I am facing the following error :

Environment :
Python - 3.6.9
Torchvision version - 0.2.0

Can you please tell me how to resolve this ?

===> Setting ...... ../models/exists using device ID: [0] ===> Building model Single-object mode save root = ../predictions/RANet_Video_17test_dev Change to multi-object mode Traceback (most recent call last): File "RANet.py", line 104, in <module> predict_SVOS(params='RANet_video_multi.pth', dataset='17test_dev', save_root='../predictions/RANet_Video_17test_dev') File "RANet.py", line 83, in predict_SVOS checkpoint_load(opt.workfolder + params, model) File "/home/swain/Documents/VOS/Methods/RANet/codes/RANet_lib/RANet_lib.py", line 38, in checkpoint_load checkpoint = torch.load(fpath) File "/home/swain/anaconda3/envs/vos/lib/python3.6/site-packages/torch/serialization.py", line 426, in load return _load(f, map_location, pickle_module, **pickle_load_args) File "/home/swain/anaconda3/envs/vos/lib/python3.6/site-packages/torch/serialization.py", line 603, in _load magic_number = pickle_module.load(f, **pickle_load_args) _pickle.UnpicklingError: invalid load key, '<'.

hello, i'm really interest in your work, is it convenient for you to provide the train code?

Thank you for your work, but when can the source code be released?

Hi！
I run this code on 1 GPU(Nvidia 1080ti). And I think your code every time process 1 to 4 frames, ringt?But when I compute the forward pass(only time between "output=model(*input)"), it shows about 230ms when batch =4. It equals about 60ms a frame. It's different from your paper... Am I wrong?

Here is my code(I add time code in RANet_lib.py):

  ` inputs = [torch.cat(Img)[index_select], torch.cat(KFea)[index_select], torch.cat(KMsk)[index_select], torch.cat(PMsk)[index_select]]

    torch.cuda.synchronize()

    stime = time.time()

    outputs, _ = model(*inputs)

    torch.cuda.synchronize()

    etime = time.time()

    model_time = (etime-stime)*1000

    print('time:',model_time)`