FPENet is difficult to reproduce
changwenkai101 opened this issue · comments
Hi, this is a good project.
I tried it, and the overall installation and training were very simple and straightforward. I experimented with FPENet, but the final result was larger than the original one. specifically:
I modified the hyperparameters to the original:
model = FPENet
dataset = cityscapes
input_size = 512, 1024
classes = 19
train_type = train
max_epochs = 400
lr_schedule = poly
The loss used CrossEntropy2d:
That is, line 133 of train: criteria = CrossEntropyLoss2dLabelSmooth (weight = weight, ignore_label = ignore_label), changed to: CrossEntropyLoss2d function, while other settings remain unchanged, using an RTX2080Ti GPU, training for 9h
But in the end, mIOU is only 46% on the val set, and the original effect is 60–70% on the test set, but I feel that there should not be such a big gap between val and test. I checked some output and noticed that the model parameters printed by train.py were 0.12M, but the original model was 0.4M. At first, I thought the model was wrong, but after checking the paper, I felt that your implementation was correct. Then I used torchsummary in the model to see that the model was 0.44M, so I didn't know what went wrong.
Maybe FPENet itself is difficult to reproduce? (Although this is common in AI papers).
Has anyone used this project to reproduce and roughly achieve the effect of an original model? Can you discuss and share the parameters and strategies set?
I also encounter this problem，FPENet result is not so good
@xiaoyufenfei Have other models been tested? Are there results that are close to or consistent with the description of the original article?
It's better to regenerate “camvid_inform.pkl” or "cityscapes_inform.pkl" according to your local dataset
So, I want to know you detail suggestion ?
So, I want to know you detail suggestion ?
I'm not sure the file "cityscapes_inform.pkl" is correct or not, it got "data['classWeights']: [ 1.4705521 9.505282 10.492059 10.492059 10.492059 10.492059
10.492059 10.492059 10.492059 10.492059 10.492059 10.492059
10.492059 10.492059 10.492059 10.492059 10.492059 10.492059
5.131664 ]"， and i got “data['classWeights']: [ 2.5959933 6.7415504 3.5354059 9.8663225 9.690899 9.369352
10.289121 9.953208 4.3097677 9.490387 7.674431 9.396905
10.347791 6.3927646 10.226669 10.241062 10.280587 10.396974
10.055647 ]”
Also, the code:
dataset_list = os.path.join(dataset, '_trainval_list.txt')
maybe should be:
dataset_list = dataset + '_trainval_list.txt'
in dataset_builder.py on line 10
Is there any difference ? you can try it, I want to know the result.