Marc Dietrichstein

Hey @zlin3000, I've put a lot of time into investigating this issue to no avail... @delta9 and you might be right in suspecting different `resize` implementations. Another reason might be...

@hristorv Not yet, I'm afraid

I have fixed a bug in the model definition (e1ada8d326d6f741962e3c6104c4d3e664d7f3a6) which definitely corrupted some classifications. It would be awesome if some of you could run your checks again and let...

Hey @delta9 Thanks for your help! > First I thought it had something to do with the image size but sadly it's all over the place. I've found out that...

I've spent some more time on this and have identified two serious problems: __Padding issues when doing pooling/convolutions__ This was due to the aforementioned different padding approaches between caffe and...

> So I would need to preprocess the images using the yahoo image loader and then send over the data for prediction - if I want to use it with...

Hey @waheebyaqub! Thank you so much for posting you results here. May I ask which dataset you were using for your test? I'm planning to use [this dataset](https://sites.google.com/site/pornographydatabase/) for a...

@waheebyaqub Alright, thanks! > So has this problem been solved? @liudanking If you use the yahoo image loader then yes, the issue is fixed. The tensorflow image loader still gives...

> BTW, is there any plan to give a guide for fine-tuning tensorflow-open_nsfw just like the yahoo one? Not in the near future since I'm spending most of my time...