bf-vos
bf-vos copied to clipboard
Pytorch implementation of Blazingly Fast Video Object Segmentation with Pixel-Wise Metric Learning (Chen et al)
When running the online retrieval, I see the following error (truncated for brevity): ``` DEBUG:__main__:Model initialized and moved to CUDA INFO:__main__:Loaded weights from ./bfvos/training/checkpoints/ckpt_epoch_1_batch_1800.pth File "./bfvos/retrieve.py", line 62, in batch_segment...
@braindeadpool Thanks for reproducing the algorithom. It may take several days to train the network. Could you please provide a completely trained model?
In train.py `main()` function, instead of `val_triplet_sampler = davis.TripletSampler(dataset=val_data_source, num_triplets=args.num_val_batches, randomize=True)` must be `val_triplet_sampler = davis.TripletSampler(dataset=val_data_source, num_triplets=args.val_batch_size, randomize=True)` to avoid throwing an exception due to incompatible tensor dimensions.
Hi, I'm confused about the structure of the DAVIS folder. When I run the code it keeps telling me there's no train.txt. Where should I find it? Thanks
Hey, Thanks for your wonderful work. Have you compute mIOU on DAVIS16 validation set of your final model? If yes, would you like to share it with me?