How to download images and bounding boxes from imageNet such that they have matching names?

Question

I am doing object detection for a specific class, say, chairs .

I want to download images of chairs from imageNet. I also want to download the annotation xml files (bounding boxes) from imageNet.

Both these things are provided on imageNet and I have successfully been able to download them using a tool called ImageNet_Utils

https://github.com/tzutalin/ImageNet_Utils

But the downloaded images and bounding boxes don't have matching names. So it is impossible to tell which xml file is for which image.

How do I download images and bounding boxes from imageNet such that corresponding image and annotation xml files have matching names?

Answer 1

The download image URLs page says

The URLs are listed in a single txt file, where each line contains an image ID and the original URL

Unfortunately, as of 2020-03-06, all the URL mapping files link to a Oops! The URL is not valid page. However, can however get mappings for each node individually. They are available by wnid: http://www.image-net.org/api/text/imagenet.synset.geturls.getmapping?wnid=n03273913

A bounding box annotation file will contain this element.

    <filename>n03273913_16800</filename>

The n03273913 is the synset id and the 16800 is the image id. In the synset mapping file you'll find the line

    n03273913_16800 http://farm1.static.flickr.com/186/425238103_8fe80b37de.jpg

You can download the image from that location.