image-segmentation-using-tiramisu
image-segmentation-using-tiramisu copied to clipboard
Model doesn't work well
Hello,
I was wondering if this model linked in the README is supposed to work directly with clothing images or we still need to train on top of it. I loaded the model according to the "View results" section in the Jupyter notebook, but I'm getting really bad results.
Thanks!
Thanks for asking this question. The model is provided as an example and as happens frequently in Deep Learning, it has learned well to segment certain kinds of pictures, where the garment is typically the most visible item in the foreground, and centered. to have an idea have a look at the sample images you can see in the notebook. as a result it is expected that with different image setup the provided network is not going to work well. can you please provide us an example of the images where the network does not work well? thanks!