GMFlowNet

GMFlowNet copied to clipboard

the model will converge faster than RAFT?

2

Since the model is initialized by the coarse flow instead of zero flow, will it converge faster than RAFT? Maybe it will get good performance with fewer training steps? thank...

Lyan-ing

Reproduction issues

5

Nice Work！！ Thank you very much for your work! However, when I reproduce your code, I never reach the metric of your paper, is there any training trick ? Looking...

xiewende

About resume training

5

if args.restore_ckpt is not None: strStep = os.path.split(args.restore_ckpt)[-1].split('_')[0] total_steps = int(strStep) if strStep.isdigit() else 0 else: total_steps = 0 Hi, I found this code in train.py. The class OneCycleLR need...

Eryo-iPython

About 4D Cost Volume Visualization

1

Nice Works ! I tried to visualize 4D cost volume by myself, but I can't achieve the appearance in your paper, can you provide your visualization code? Looking forward to...

xiewende