NHW2017

Results 17 comments of NHW2017

> > Maybe you should update the version. The torch and CUDA used in this version are too old, and there are many problems in the implementation. > > have...

> > Maybe you should update the version. The torch and CUDA used in this version are too old, and there are many problems in the implementation. > > have...

> As there are variances of the novel samples of phase 2, we evaluate the model five times to get the average results as our paper. Because there are no...

> A CUDA version of 9.0 or 9.2 works for this code. The problem encountered by CUDA10.0 remains unsolved. @Lemonqinnn @tooHotSpot I tried to run the program under the environment...

> I found that in the metadata.py, if phase==2, shots=shots*3, is this a bug? I think this will lead to an unfair performance comparison. Sorry to interrupt you, I have...

thank you for your reply! In fact, I have tried many environment configurations according to the "ReadMe" of the two programs "faster-rcnn.pytorc" and "MetaRCNN", but there is always an error....

> hi! do you know how 'nongt_dim' works? > https://github.com/msracver/Relation-Networks-for-Object-Detection/blob/e83e911d828e3c86624ce0aeb8d742d5ee67d5ba/relation_rcnn/symbols/resnet_v1_101_rcnn_attention_1024_pairwise_position_multi_head_16_learn_nms.py#L85 > > > for example, nongt_dim = 2000(rois per image), when train, there is 2 images per batch, so 2000*2...

> Hi @NHW2017 > > Yes, the environment configuration is important, especially use the correct PyTorch version and CUDA version. > > Two commands might be useful when installing the...

> Actually, the training time is correct. The base class training on VOC also took me around 30 hours. > To accelerate, one solution would be implementing parallel training with...

> The base training on COCO would take ~10 days, which is unfortunately too long with only one GPU of 12G. > A work around is to increase the batch...