-
I clone the project and running the code with the same hyperparameters in the paper. The recall@1 on Cub200-2011 is 63.79%, which is far from the result in paper 80.5. P.S. the model is DREML(R,12,48)…
-
Hi, i have clone your code, and the performance on CUB is
R@1: 51.5
R@2: 63.9
R@4: 73.7
R@8: 82.3
but the R@1 on Cars196 Up to 60.0
the paper result: R@1 on Cars196 is 73.2
-
Hi, @dichotomies
I just tried "python train.py" on cub200, but the performance was not as good as yours.
So, I'd like to know if I have missed some details.
Thanks!
------------------------------…
slpal updated
5 years ago
-
could you tell me your cars196 args.json
this is my args.json
{
"batch_k": 4,
"batch_p": 32,
"checkpoint_frequency": 800,
"crop_augment": false,
"decay_start_iteration": 2000,
"de…
-
Hi,I'm a new learner.I do not understand how to prepare dataset for train,eventhough I've read your instruction.I download market1501 dataset,and I create a folder named images to include all imgaes o…
-
Hi there,
You posted a comment on https://github.com/bnulihaixia/Deep_metric/issues/5#issuecomment-410479866 and I checked out your work here, nice job!
During my research around this loss, I fo…
-
I attempted to introduce a new .py file in dataflow/dataset folder to make it convenient to load another new dataset (e.g., Cars196). But after that, my tensorpack.dataflow.dataset cannot find the Car…
-
I can use ftp://cs.stanford.edu/cs/cvgl/ to download CARS196.zip
-
Hi, I have a similar implementation in mxnet but after the first epoch both the test and train accuracy goes down (though loss as well) instead of going up. I tried your code as well with my dataset a…
-
```
cluster_20181113035415-806898
{
"image": {
"width": 256,
"height": 256,
"channel": 3,
"random_crop": {
"width": 224,
"height": …