cspearl
cspearl
> Interesting. I have DenseCap fully working on my Ubunutu VM which is version 20.04. I'll look into it tomorrow and ask any further questions if they come up. Thanks....
> Would you mind proving the implementation of your dataset and the documentation link you referred to? I used this instructions while implementing my dataset - https://github.com/MengzhangLI/mmsegmentation/blob/add_doc_customization_dataset/docs/en/tutorials/customize_datasets.md This is the...
> I think you might print `kwargs` after `def __init__(self, **kwargs)` before `super` to check the `times` in `kwargs` {'times': 40000, ...} this gets printed as the first key-value pair...
> Do you use `RepeatDataet` in dataset config? > > https://github.com/open-mmlab/mmsegmentation/blob/4eaa8e69191cc293b64dafe47f1f88a7d468c93c/mmseg/datasets/dataset_wrappers.py#L176 No I have not used RepeatDataset in config My config is - dataset_type = 'AmazonDataset' data_root = 'data/amazon' img_norm_cfg...
Yes the training part is now working thanks for the help!
> 在使用dataloader时候，输入图片需要变成相同的尺寸才能压入一个batch中进行训练，如果此过程是训练过程，建议检查数据增强中是否有reshape到统一的大小的操作。 Yes it is a training process but all my images are already of the same size - 512x512 RGB images and labels are 512x512 greyscale
While studying the code I found out for the databases tf.decode_jpg or png is used but my dataset has images of type .tif (1024*1024) . I saw that there is...
Do you know exactly what tf.decode does? Then probably I could use some other function for tiff images?
Yes i checked version 1 of tf does not support tif neither does v2. But there is an option in experimental tf-io as I had said but I will try...
> Hi, @cspearl, could you try to set `reduce_zero_label = False`, and `num_classes = 1` and ensure your annotation files only have label `0` and label `1`? Yes I tried...