Guo Rui
Guo Rui
Thanks!
It runs well.
Are there parameters need to be changed? before  after  Anyway, thank you very much for sharing the cocolor and mosaicing code
In the sigmoid_cross_entropy_loss_layer.cpp file, target means the label data block and there are two branches when computing the loss. So the label should be converted to label[(label>0)]=1. Am I right?...
Rethinking Atrous Convolution for Semantic Image Segmentation. Liang-Chieh Chen George Papandreou Florian Schroff Hartwig Adam
image_dir, key_frame, output_dir
运行到 return &Jieba{ C.NewJieba( dpath, hpath, upath, ipath, spath, ), } 程序直接退出了
This problem occurs when I use video_take_snapshot method loading http-flv stream. No this problem in the VLC player application loading http-flv stream.
ip都被封了?
一个都扫描不出来了
InvalidArgumentError (see above for traceback): ConcatOp : Ranks of all input tensors should match
25_Soldier_Patrol_Soldier_Patrol_25_457.jpg Traceback (most recent call last): File "./train_net.py", line 134, in restore=bool(int(args.restore))) File "./lib/tiny/train.py", line 338, in train_net sw.train_model(sess, epochs, restore=restore) File "./lib/tiny/train.py", line 180, in train_model score_cls_np, score_reg_np, prob_cls_np...
I did not find any config file to set it.