-
tensorflow.python.framework.errors_impl.InvalidArgumentError: indices[176] = 2530 is not in [0, 2520)
[[{{node InceptionV3/bilinear_attention_pooling/logits/Gather/GatherV2}}]]
I got this Error …
-
Great repo and well written blog! I have one question: RoiPoolConv uses tf.image.resize_images, which performs a bilinear interpolation, not a max of the pixels (unless I am misinterpreting its use he…
-
Hi @gy20073 ,I can only find the Compact Bilinear Pooling applying in VGG16 network. So have you ever tried other networks. I try to use ResNet18 with learning rate of conv bn and scale layer setting …
-
尝试将模型从pytorch-》onnx-》ncnn,转换期间由于各框架支持算子的限制,将adaptiveAvepooling换成普通的pooling,F.interplate的插值方式从bilinear换成了nearest,经过训练在python上能得到非常好的分割结果,但是转换到ncnn进行C++推理后,同样的图片分割结果很差,请问有人转PSPnet碰到过这钟问题吗?
-
您好,我想请教您几个问题?
(1)在resnet50里, 代码实现中attemtion map 的数量为32,是每一个attention map 和原来的最后一个bottleneck里的卷积层特征相乘。我想问的是attention map是如何生成。
(2)能否采用像imporved bilinear pooling的方法改良整个代码。
-
Sorry to disturb you again,How to solve this problem?
![_20181025170648](https://user-images.githubusercontent.com/36980516/47494084-7892c480-d883-11e8-9a2b-10606c7c599f.png)
-
Hi @gy20073 ,
I am trying to test the output of the compact bilinear layer but can't get the result right. I have used this for training and the result is very good. This means the layer should wor…
-
训练了53个epoch,测试准确度还是只有0.39%,第1个epoch的时候准确度就是0.4多。说明真个训练网络都没有学习到。想问下我们这工程是论文官方的工程代码吗?是否能给出我们训练的结果?
训练数据使用bird,过程数据如下:
epoch | train_acc | train_loss
-- | -- | --
3 | 0.3504 | 5.3009
4 | 0.317 | 5…
-
1. image size problems in training:
When the image size in my own dataset is small (e.g. h=200, w=80), when I train the model, how should I set the INPUT_SIZE?
INPUT_SIZE='720,720'
or
INPUT_SIZ…
-
Hello, what is the function of the parameter sum_pooling in the code? If true seems to be a summation operation, is there any difference in principle if no summation is made? I don't quite understan…