andyhahaha
andyhahaha
I change the backbone to PeleeNet and train with 4 gpu. But feat_id will have some elements are nan and cause the pooled feature lost some feature and dimension mismatch...
The paper said that NSGA-Net + macro search space takes 8 GPU-days NSGA-Net (6 @ 560) + cutout takes 4 GPU-days But I run this searching code on Tesla P100-PCIE-16GB,...
I use this code to train segnet on camvid, but the results can't reach the paper performance. I found that you didn't use median-frequency balancing to training. Moreover, unlabelled class...
I read the R-FCN paper. After the Resnet101 they append a randomly initialized 1024-d 1x1 convolution layer to reducing dimension. However, your implementation append the a 1024-d "3x3" dilation="6" convolution...
Hi, I run dff_rfcn_test.py to test model's mAP. However after it print the mAP, the program hang and never stop. I find out there is multiprocessing.pool to do parallel evaluation....
Hi all, I try to apply other flownet to DFF. I need to confirm the image preprocess is the same. However, I found the image is convert from BGR to...
I'm trying convert RefineDet to caffe model. RefineDet only contain regular CNN module(conv2d, convtranspose2d, maxpool, batchnorm). https://github.com/luuuyi/RefineDet.PyTorch However, pytorch2caffe failed at add operation as follow:  This add operation is...
I can't understand the loss function of FBnet. There are cross-entropy loss and latency loss. Usually multiple loss weighted sum together become total loss. Why FBnet multiply them together? Besides,...
I found that the label of Seq05VD_f02610 in test data is wrong. Do anyone notice this? Why that happened?
Do anyone successfully train on coco and reach the paper accuracy? (refinedet 320 VGG: 29.4) Can somebody provide the coco evaluation code? I want to train coco myself but I...