Grant Van Horn

Started a tutorial here: https://github.com/visipedia/tf_classification/wiki/CUB-200-Image-Classification

My final accuracy using only the image is around 82 - 83%. I haven't been able to reproduce Krause's results, which actually use the Inception-v2 model. I've talked with Krause...

You are right, it was the Spatial Transformer paper that used inception-v2. Krause used inception-v3. Here are some notes from an email exchange with Krause: > From my notes an...

Yes, I have tried, but the numbers were still off. A few of the lab mates have also tried unsuccessfully to reproduce those numbers. On Tue, Dec 5, 2017 at...

I've tried to reproduce your error by messing with my mongodb settings, but I haven't had luck. So a few questions: 1. Do you have mongodb running? 2. Have you...

Sure, so to turn of the augmentation you'll need your `IMAGE_PROCESSING` section of your config script to look something like: ``` IMAGE_PROCESSING : { # All images will be resized...

Ah, I see. But in TensorBoard are you seeing the same image for all of the summaries? Hopefully that is the case. The summary images will still be produced (and...

Cool. The only image that is being queued for training is the last summary image (`3.final_distorted_image`). The other image summaries are a convenience to see the intermediate augmentations that were...

Yeah, you can keep the same `IMAGE_PROCESSING` configs for your test situation. For the batch size, what I'll do is use the following code snippet (taken from [here](https://stackoverflow.com/questions/6800193/what-is-the-most-efficient-way-of-finding-all-the-factors-of-a-number-in-python)) to find...

The `ACCURACY_AT_K_METRIC` is way of specifying the accuracy measurement you would like rendered on TensorBoard. Specifying `ACCURACY_AT_K_METRIC : [3, 5]` means that you would like accuracy at 3 and accuracy...