Update annotations on data_loader

[danielsyahputra]

:camera: Screenshots

:page_facing_up: Context

I realized in your code before, that you just assign '1' as the labels for each object. This is proved by creating a tensor of ones for labels like this labels = torch.ones((num_objs,), dtype=torch.int64). When I tried my model to do inference on my sample image, I got the labels '1' for each object and then I realized there was something wrong with the dataset.

:pencil: Changes

I just add a little bit of code on your custom Cppe dataset in torch.py. Now, the labels not only '1' for each object in an image, but also have a correspondence with each object based on your dataset.

:paperclip: Related PR

:no_entry_sign: Breaking

None so far.

:hammer_and_wrench: How to test

:stopwatch: Next steps