janesjanes / sketchy Goto Github PK

Hello, excuse me, can you please share the sketchy dataset? Thank you!

Thank you for the great code base!

I was hoping to replicate your results, but found that the image lists in your train prototxt are not in the repo anywhere: https://github.com/janesjanes/sketchy/blob/master/training/Triplet_googlenet_train_test.prototxt - would you mind sharing these? Thank you!

Can you please kindly provide one ? Thanks

https://goo.gl/3EZgxp

Is it possible to update it/upload a new one?

Hi, the url of sketch model is not accessible，https://goo.gl/cqXm7F

It shows below:

Hi! I'm working with this code for another project. I've pushed some modified notebooks into my repo if you're curious https://github.com/kylemcdonald/sketchy/tree/master/code

Including a example for uploading sketches from the browser to a backend that processes them and returns similar images (Server.ipynb).

I'm interested in re-training the network on a synthetic dataset, where we run some smart edge detection to generate sketches from photos. Is it possible to share the solver and training files (or code) that were used to train the network?

Thanks!

Hi,

there are several folders of the rendered 256x256 images, what's the difference?

256x256/
|-- photo
|   |-- tx_000000000000
|   `-- tx_000100000000
`-- sketch
    |-- tx_000000000000
    |-- tx_000000000010
    |-- tx_000000000110
    |-- tx_000000001010
    |-- tx_000000001110
    `-- tx_000100000000

I am researching a problem of the image synthesis. I want to run the code of the Scribbler but the link of the homepage has expired. Can you provide me the new link of the Scribbler? Thanks!

I was interested in exploring sketchy's Triplet Loss + Classification Loss network, though having a little trouble dissecting the different losses in the network, and what weighting you would have used for each of these for the model described in your paper. If I try fine tuning with your model and using the sketchy db w/ 125 categories and removing the invalid data from your shared annotations lists, using what I think are image lists that you might have used, the Euclidean loss starts to blow up. Is this expected behavior, or did you weight this loss higher than 0 to ensure it did not blow up?

Is this network solver/prototxt exactly the same as what you used for the reported Triplet Loss + Classification Loss network in your paper: https://github.com/janesjanes/sketchy/blob/master/training/Triplet_googlenet_train_test.prototxt

It looks like the softmax loss for the sketch was set to 10, negative image to 0, and positive to 10.
And for triplet loss, you had three separate ones set to 0.3, 0.3 and 1 (not sure I understand what each of these three separate losses are?), and for contrastive loss, all are set to 0. Finally, there is also a euclidean loss which is set to 0. Any advice/guidance would be much appreciated!