Whitebear

It's normal behavior when the model is overfitting. (https://www.dataquest.io/blog/learning-curves-machine-learning/) If you split the dataset with training, validation and test set, you can apply early stopping technique. https://page.mi.fu-berlin.de/prechelt/Biblio/stop_tricks1997.pdf To avoid overfitting,...

Which kind of model do you want to run? Input image size can be divided by 32 for some models. You have to read the paper and figure out the...

That error you have is about GPU memory. You don't have enough GPU memory to feed-forward that size of the image. I think you have to change the larger GPU...

There is no pre-trained model provided in this project. You have to train the model by yourself.

How did you set about the cropsize?

Did you check the image is loaded properly?

You should definitely provide more information on the issue. such as... What kind of model and frontend did you use? Did you modify the code? Which dataset did you use?

you can check the code of deeplab v3+. https://github.com/rishizek/tensorflow-deeplab-v3-plus/blob/master/utils/preprocessing.py#L115 That function will return re-scaled image with the scale from min_scale to max_scale. Moreover, if you get ``` end_points['pool3'] ``` ,...

Hi, 1st answer: you can remove the background if you could segment the person correctly. You can get the binary segmentation map (person or background) and take the bit-wise and....

What do you mean "correctly segmented"? If the output image at the prediction stage is totally wrong, Did you try to re-write ``` update_ops = tf.get_collection(tf.GraphKeys.UPDATE_OPS) with tf.control_dependencies(update_ops): ``` before...