question about test model on another dataset

[cena001plus]

Thanks to the author's code contribution, when I use the code, I found the following problems:

1. The model obtained after the code is run, only one test picture is saved, not all the pictures in the test directory;
2. This code is 12 categories, the value of num_classes is different in different places, 11, 12, 13, what does this mean? If I am modified to a second classification, how do I modify these values?
3. Using quene's batch data method, what is the model input node for model?
4. When the ckpt model is transferred to the pb model, it always goes wrong. The converted pb model is blank, but I have tried different output nodes;

5.The modified code 12 is classified into 2 categories, and the training precision with only one classification is high, and the other is close to 0;
6. When testing the model with the command, the following error occurs.

>python main.py --testing=./logs/model.ckpt-99 -log_dir=./logs/ --test_dir=./SegNet/myData_3/test.txt --batch_size=5 --save_image=True
.............................................
........................................
..........................................

ValueError: Cannot feed value of shape (1, 200, 200, 3, 1) for Tensor 'Placeholder_1:0', which has shape '(1, 200, 200, 1)'

I have modified the data format of placehold, the model can be trained, the data path is ok, but when this problem occurs, I don't understand what the problem is. Does anyone know what the problem is?

[cena001plus]

ValueError: Cannot feed value of shape (1, 200, 200, 4, 1) for Tensor 'Placeholder_1:0', which has shape '(1, 200, 200, 1)'

what that mean？

[duocang]

@cena001plus I am new(2 months) too. But it is clear that your input image is not in the right size. I

I have no clues why your input has four dimensions(let's call it dimension.). What is the 4 in (1, 200, 200, 4, 1)?

[song248]

I have same problem. Do you solve that error?