Stanford Cars Dataset Annotations missed

Question

I need help, I am currently working a neural network for object detection. I'm using standford cars dataset from Kaggle as my training and testing dataset. but I have a problem in the annotations. the ground truth boxes are not covering the entire cars. is it like that? or I read it in wrong ways? this is the link if you need to check it. Thanks for helping me. Have a nice day :)

[Stanford Cars Dataset] https://www.kaggle.com/jessicali9530/stanford-cars-dataset

Hi! Welcome to StackOverflow! I would suggest you to be more specific in what you need, and also to post whatever attempt you have for what you are trying to achieve. See also [How do I ask a good question?](https://stackoverflow.com/help/how-to-ask) and [How to create a Minimal, Complete, and Verifiable example](https://stackoverflow.com/help/mcve). — norok2, Sep 26 '18 at 12:03

score 0 · Accepted Answer · answered Sep 27 '18 at 17:55

0

You might find the paper that discusses how the dataset was put together helpful. It was intended for 3D rather than 2D object detection, which could be why the ground truth looks slightly different than you'd expect for 2D modelling.

Link to paper: http://vision.stanford.edu/pdf/3drr13.pdf :)

answered Sep 27 '18 at 17:55

Rachael Tatman

841
7
6

thanks your answer really helping me. so that's why I can't fully utilized the annotations in my 2D object detection. – Kevin A.S. Sep 28 '18 at 02:42
then how should we use the annotation if we want to proceed with 2D recognition?? – Vincent Tang Jun 03 '19 at 17:40

Stanford Cars Dataset Annotations missed

1 Answers1