Nicolai Wojke

Empty classes will occupy _some_ space in the feature space and samples which fall within this area are going to receive a penalty. Since the scale parameter is shared between...

The checkpoint file that ships with the deep sort repository has some minor differences compared to the version in this repository. The deep sort repository uses a frozen inference graph...

Hi ngingihy, it is correct that you need to to train on a custom dataset for your application. The steps are roughly (1) get dataset, (2) write a dataset adapter...

Great to hear that you made progress. I haven't seen the error before. It would be great if you tell me where it occurs.

Can you be more specific? If you follow the calls to this function in code, you see that they are called for example [here](https://github.com/nwojke/cosine_metric_learning/blob/master/train_market1501.py#L20). The purpose of these functions is...

That sounds like overfitting and you probably need to change hyperparameters. Maybe it is a good idea to start with the weight decay of the scale parameter. It is currently...

The important line seems to be the following, indicating a shape mismatch for the input images. I have trained on VeRi yesterday without issues. Which TensorFlow version are you using?...

We are currently not making the VeRi model public because the dataset requires signing an agreement form. The code that we provide can be used to train on VeRi without...

Are you sure that you didn't make any changes to the code in branch ``feature/veri_dataset``? I haven't made any changes and trained on VeRi when I made my comment on...

Looks like there is something wrong with the data. My guess is that the error occurs during batch sampling. Have a look at issue #9. The training script prints the...