PSPNet-tensorflow

PSPNet-tensorflow copied to clipboard

I'd like to train on [Kitti dataset](http://www.cvlibs.net/datasets/kitti/eval_semantics.php), which is of size **1242 x 375.** Would this work with the current pool size? **Update:** Tried to train with the default parameters...

ghost

i trained the new model,and modified the Num_class to 3,original 19, however, i got the results img fullfilled with pink color after i ran the inference, i don't know why?

10

as i mentioned above, i retrained the model with the input imgs 1280x720, and classified into 3 NUMS_classes, trained models based on the pretrained models. finally, it generated new *.cptk...

sainttelant

OOM when allocating tensor with shape

4

Hi, thank you very much for contributing this codes. However, when I evaluate with ade20k dataset, I got the error: "OOM when allocating tensor with shape" I'm using two Tesla...

TingZhang08

i trained the downloaded pspnet model 1 epoch over all the ade20k training dataset and i got miou=0.274 because of the small batch size used (batchsize=1) but when i used...

AyaMohamedS

Training with a different number of classes

5

Thanks for this great repository! I am trying to train the model on my own dataset which has only 5 labels. When I try to train with loaded checkpoints from...

anny123123

I wanna test my own picture, however ,when I use python inference.py --img-path=./input/indoor -1.jpg --dataset ade20k, it have some wrong ,I really don't know why?

2

could anyone help me? python inference.py --img-path=./input/indoor_1.jpg --dataset ade20k successful load img: ./input/indoor_1.jpg WARNING:tensorflow:From /home/nk/PSPNet-tensorflow/tools.py:84: calling expand_dims (from tensorflow.python.ops.array_ops) with dim is deprecated and will be removed in a future...

Kedron007

Some questions about the loss

4

When I use this code to train the cityscape as the paper say. But I found that the loss is always at about 2~~3. I want to know if this...

FeiWard

Initial model

2

@hellochick Nice work, I found that there is no initial model in your checkpoint Google Driver folder, could you share your initial model which you train your code? Thank you...

FengLoveBella

Training crashes on Cityscapes

2

I'm training PSPNet using the train.py script provided - i've tried running it on GTX1080 and TitanX . It always crashes after about 500 steps. Log below: ``` step 590...

kshitijagrwl

Hi, Thanks for your works! @hellochick I get a problem when I use your model and evaluate.py to evaluate my image in cityscape. If I use 2048*1024 resolution, it can...

Leequens