# Downloading ImageNet

**URL:** <https://discuss.pytorch.org/t/downloading-imagenet/122338>\
**Category:** Uncategorized\
**Created:** [May 25, 2021, 5:19pm UTC](https://discuss.pytorch.org/t/downloading-imagenet/122338 "2021-05-25T17:19:22Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![julianolm](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/julianolm/32/35135_2.png) [@julianolm](https://discuss.pytorch.org/u/julianolm)\
**Post date:** [May 25, 2021, 5:19pm UTC](https://discuss.pytorch.org/t/downloading-imagenet/122338/1 "2021-05-25T17:19:22Z")

</div>

Hello, I am developing a model to apply on FMD ([Flickr Material Database](https://people.csail.mit.edu/lavanya/fmd.html)), but training on that same database just lead to 30% accuracy. Now I’m gonna pre-train the model on ImageNet, but don’t know how to do it. I haven’t yet even discovered how to download it in a simple way. How should I do it?

Also, since don’t have GPUs I am using Colab, wich has a small storage (64GB) in comparison to ImageNet’s size (think 150GB zipped). So I think my way is to upload the dataset to my edu drive account, wich has unlimited storage (as for now).

Thank you!

---

<div class="post-metadata">

**Author:** ![eqy](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/eqy/32/46233_2.png) [@eqy](https://discuss.pytorch.org/u/eqy)\
**Post date:** [May 26, 2021, 6:08am UTC](https://discuss.pytorch.org/t/downloading-imagenet/122338/2 "2021-05-26T06:08:23Z")

</div>

If you are just prototyping you might be able to just use a pretrained torchvision model:  
[https://pytorch.org/vision/stable/models.html](https://pytorch.org/vision/stable/models.html)

Downloading from the [original website](https://image-net.org) is often tricky, but the [academic torrents](https://academictorrents.com/collection/imagenet) version works well if you have a way to mount or access \> 150GB of storage.