Daily Archives: December 22, 2021

minDALL-E creates images based on text input

Posted on by

minDALL-E on Conceptual Captions

minDALL-E, named after minGPT, is a 1.3B text-to-image generation model trained on 14 million image-text pairs for non-commercial purposes.

a painting of a bird in the style of asian painting
a photo of san francisco's golden gate bridge in black and white tone

Environment Setup

  • Basic setup
PyTorch == 1.8.0
CUDA >= 10.1
  • Other packages
pip install -r requirements.txt

Model Checkpoint

  • Model structure (two-stage autoregressive model)
    • Stage1: Unlike the original DALL-E [1], we replace Discrete VAE with VQGAN [2] to generate high-quality samples effectively. We slightly fine-tune vqgan_imagenet_f16_16384, provided by the official VQGAN repository, on FFHQ [3] as well as ImageNet.
    • Stage2: We train our 1.3B transformer from scratch on 14 million image-text pairs from CC3M [4] and CC12M [5]. For the more detailed model spec, please see configs/dalle-1.3B.yaml.
  • You can download the pretrained models including the tokenizer from this link. This will require about 5GB space.

Sampling

  • Given a text prompt, the code snippet below generates candidate images and re-ranks them using OpenAI’s CLIP [6].
  • This has been tested under a single V100 of 32GB memory. In the case of using GPUs with limited memory, please lower down num_candidates to avoid OOM.

[…]

Samples (Top-K=256, Temperature=1.0)

  • “a painting of a {cat, dog} with sunglasses in the frame”
  • “a large {pink, black} elephant walking on the beach”
  • “Eiffel tower on a {desert, mountain}”

More

There’s dalle-mini, a colab where you can run it to test it

This App Will Tell Android Users If an AirTag Is Tracking Them

Posted on by

Apple’s AirTags and Find My service can be helpful for finding things you lose—but they also introduce a big privacy problem. While those of us on iOS have had some tools for fighting those issues, Apple left those of us on Android without much to work with. A new Android AirTag finder app finally addresses some of those concerns.

How AirTags work

[…]

The Find My network employs the passive use of hundreds of millions of Apple devices to help expand your search. That way, you can locate your lost items even if they’re too far away for traditional wireless tracking. Your lost AirTag may be out of your own phone’s Bluetooth range, but it may not be far from another Apple device.

[…]

The Tracker Detect app comes out of a need for better security in the Find My network. Having such a wide network to track a tiny, easy-to-miss device could make it easy for someone to use AirTags to track someone.

People pointed out this vulnerability pretty soon after Apple announced the AirTags. With more than 113 million iPhones in the U.S., not to mention other Apple devices, the Find My network could be one of the widest tracking systems available. A device as small and easy-to-use as an AirTag on that network could make stalking easier than ever.

That said, Apple has a built-in feature designed to prevent tracking. If your iPhone senses that a strange AirTag, separated from its owner, is following you, it will send you an alert. If that AirTag is not found, it will start to make a sound anywhere from 8 to 24 hours after being separated from its owner.

However, Android users haven’t had these protections. That’s where Tracker Detect comes in; with this new Android AirTag app, you can scan the area to see if anyone may be tracking your location with an AirTag or other Find My-enabled accessory.

How to use Tracker Detect

If you’re concerned about people tracking you, download the Tracker Detect app from the Google Play Store. You don’t need an Apple account or any Apple devices to use it.

The app won’t scan automatically, so you’ll have to look for devices manually. To do that, open the app and tap Scan. Apple says it may take up to 15 minutes to find an AirTag that’s separated from its owner. You can tap Stop Scanning to end the search if you feel safe, and if the app detects something, it will mark it as Unknown AirTag.

Once the app has detected an AirTag, you can have it play a sound through the tag for up to ten minutes to help you find it. When you find the AirTag, you can scan it with an NFC reader to learn more about it.

[…]

 

Source: This App Will Tell Android Users If an AirTag Is Tracking Them