ypwhs / dogbreed_gluon Goto Github PK

I am trying to run the model, but there's a error in this line:
features = np.concatenate([features_resnet, features_densenet, features_inception], axis=-1)
Error is caused by the inconsistent dimensions.

Then I check the shape of features_resnet, features_densenet, features_inception. Find that their shapes are not the same. I pretrain the features using the preprocessing file.

my code went error when running for X,y in tqdm(data)
I have installed all dependence. But went error like following
terminate called after throwing an instance of 'std::bad_alloc' what(): std::bad_alloc Aborted (core dumped)

Have you ever met this error before?Thanks.

使用你的代码运行时，

data_iter_224 = gluon.data.DataLoader(gluon.data.ArrayDataset(X_224), batch_size=batch_size)

查看资料，关于gluon.data.ArrayDataset需要两个参数 ?请问是什么问题

mxnet.gluon.data.ArrayDataset(data, label) link

当我运行get_features.ipynb中的下面这一行:
features_vgg, labels = get_features(models.vgg16_bn(pretrained=True, ctx=ctx), data)
程序会运行很久之后便会出现错误BadZipFile File is not a zip file。程序运行时显示是从一个网址下载vgg16_bn-7f01cf05.zip文件，点开这个网址提示说这个xml网址并未包含关联信息，请问这个错误应该如何解决呢？

当我运行get_features.ipynb中的下面这一行:
features_vgg, labels = get_features(models.vgg16_bn(pretrained=True, ctx=ctx), data)

便会出现错误:The kernel appears to have died. It will restart automatically.

然后kernel 重启, 反复循环. Jupyter notebook错误日志:
terminate called after throwing an instance of 'std::bad_alloc'
what(): std::bad_alloc

应该不是GPU内存不足,我把前面一行:data = gluon.data.DataLoader(imgs, 64) 中的bach_size从64改成2还是报同样的错误. 系统内存也还没有用满,大概4G的样子. 请问这是怎么回事?谢谢

我的运行环境:
ubuntu16.04
mxnet1.0
硬件配置:
内存: 16G
gpu: gtx1080

我在跑get_features_v3的时候跑第二段的如下第二行代码时候
img_224 = ((cv2.resize(img, (224, 224))[:,:,::-1] / 255.0 - mean) / std).transpose((2, 0, 1))#此行无问题
img_299 = ((cv2.resize(img, (299, 299))[:,:,::-1] / 255.0 - mean) / std).transpose((2, 0, 1))
下方的进度条跑到78%之后就会报错Canceled future for execute_request message before replies were done
我搜索了一下说可能是opencv2的版本问题，我更换了好几个版本，有时候卡到81% 有时候88% 但是就没有跑完的。我也尽量把我的版本更新到和您一样了，如
mxnet-cu101=1.5.0
numpy=1.13.3（sciipy=1.4.0）
tqdm=4.11.2
sklearn不知道为什么只有0.0.post1
如果您还记得的话恳请帮个忙，谢谢大佬

ps：感觉299是为了inceptionv3专门处理的，这条会报错就暂时不用这个模型跳过了