Anil Batra

hi @Flame-Master, thank you for your interest. I am unable to find the files/solution currently. However, I can definitely help you with any errors. Thanks

Hi @XinlingQiu, @kangkau Thank you for using the work to explore your research. I assume you are using the similar splits for train/valid on DeepGlobe (available in the repo: [link](https://github.com/anilbatra2185/road_connectivity/tree/master/data/deepglobe))....

Hi @kangkau, The IoU mentioned in the paper is on my splits which are available in the repo. Thanks Anil

Hi @mathfinder, thanks for working on it and migrating to version 1.4. Apologies for delayed response. In my training process miou stays around ~ 0.47-0.48 for first few epochs (~...

Hi @puyangma, Thank you for showing interest in the work. I have added initial code for pre-training stage. In Post-training step, we just need to replace the corruption input with...

Hi @linhanxiao, the details are: GPU - Nvidia 1080 with 2 GPUs Training Time: close to 2 days for 120 Epochs. I will soon upload the training models. Thanks

How do we calculate these values for other datasets e.g. Celeb or CIFAR or will it be going to same? Thanks

hi @dreamgonfly , I wonder if you tried training the model without transcripts and get the similar results in Table 2, Row#1, as this does not need any pre-training or...

I am unable to reproduce the results of Row#1 in Table 2 i.e. using only visual input without any pre-training. I am using single A100 (80 GB) GPU to run...

thanks @dreamgonfly for replying back! As I am unable to reproduce the results with only visual input, which make me think that ASR (from whisper) might not be the concern....