chengjuntao comments

Results 12 comments of


                                            chengjuntao

out of bounds for axis 2

Have this problem solved? Have you try to prune mobilenetv2? how to solve Multiple input in short?

exuse me！！！！i have question,my data is image which full of long words,just like books,but when i train model by my data,it can not detect all the words in one line.maybe it can only detect half of a line and my input_size is 768 and text_scale is 768.could you tell me where i maybe wrong?thank you so much

i am sorry , i ever thought issue can not show image.......@argman just like this image

exuse me！！！！i have question,my data is image which full of long words,just like books,but when i train model by my data,it can not detect all the words in one line.maybe it can only detect half of a line and my input_size is 768 and text_scale is 768.could you tell me where i maybe wrong?thank you so much

could you tell me why it can not detect long text? in fact, is long text regression hard to train? @argman

exuse me！！！！i have question,my data is image which full of long words,just like books,but when i train model by my data,it can not detect all the words in one line.maybe it can only detect half of a line and my input_size is 768 and text_scale is 768.could you tell me where i maybe wrong?thank you so much

yes, i see, thank you so much!!!

Some questions regarding the training

@raywzy @windj007 Hi！Have you try AOT-GAN generator to get a great result? I try to replace ffc_resnet to AOT-GAN generator which reduce some channels. The model is trained on A100,...

Some questions regarding the training

I see! Thanks for your reply! I will try again~

请问下，项目中生成数据时该如何做数据类别平衡？

我指的是字的类别比例，比如一共5990类，语料中每个字的出现的频率是不一样的，直接随机生成应该会生成更多的高频率的字吧

请问一下，下面两个损失都输针对判别器的损失，输入都是pred_fake,但是target确不一样，这样每一次优化的时候会不会造成混淆？

def compute_generator_loss(self, input_semantics, real_image): G_losses = {} fake_image = self.generate_fake(input_semantics) pred_fake, pred_real = self.discriminate(input_semantics, fake_image, real_image) G_losses['GAN'] = self.criterionGAN(pred_fake, True, for_discriminator=False) if not self.opt.no_ganFeat_loss: num_D = len(pred_fake) GAN_Feat_loss = self.FloatTensor(1).fill_(0)...

请问paper的结果是用train+val数据得出的吗？

请问下，训练时使用train+val数据，指的时val也参与训练吗？paper里的指标时跑test数据得出的吗？但是test里没有label，这个是怎么得出的呢？

Running with CPU only

hello, i try to run code on cpu, when i change code according like you, it caused like below: ![image](https://user-images.githubusercontent.com/17508662/33361228-2f1ca37c-d512-11e7-9c38-ef7b47d23e00.png) i have enough 17G RAM and i do not kmow...