🚨 AI Blitz 13 ⚡ Faces [New Challenge Launch]

AI Blitz XIII
#1

Link Preview
Link Preview2400×1256 181 KB

Each face has a story to tell. Our faces allow us to communicate without using any words. This Blitz brings you the essential computer vision AI problems around face recognition. Solving these 5 Blitz puzzles will prepare you to tackle more advanced face recognition problems in AI. ​​Backed by easy-to-understand starter kits and active support from AIcrowd Community, make your first submission in 15-minutes!

In classic Blitz tradition, leaderboard toppers and community contributors stand a chance to win from a cash prize pool of $400! What are you waiting for? Start solving AI Blitz 13 puzzles so find out what lies behind these faces.

  1. Sentiment Classification: Our eyes can identify any emotions by looking at a face, but can we train an AI model to do that?
  2. Age Prediction: Built a model to guess the approximate age of a person given their image.
  3. Mask Prediction: Can we predict the mask type and bounding box from just an image?
  4. Face Recognition: In the sea of faces, find the target face using your model.
  5. Face De-blurring: No more blurry selfies, create a model to get clear images.

:date: TIMELINE

Start Time: 4th February 2022 | 12:00 PM UTC
End Time: 25th February 2022 | 12:00 PM UTC
Duration: 21 days/3 weeks

:trophy: PRIZES

Leaderboard 1st Place: $100
Leaderboard 2nd Place: $100
Community Contribution: 2 x $100

Friendly Reminder

  1. Find teammates for this challenge over here .
  2. We have a $200 Community Contributor prize pool, click here to know more & participate.
  3. The top two entries on the leaderboard will win $100 each! Click here to check out the leaderboard.

For any other comments or queries, reply in this thread and we will get back to you!:slight_smile:

#2

Hi, the challenge sounds quite interesting! I’ve got a question about the rules: is it allowed to use pre-trained networks? For example, can I use a pre-trained ResNet (or maybe even CLIP) for feature extraction or to fine-tune its weights?

#3

Hi @notnanton

Glad to hear that you liked the challenge :slight_smile: And yes, you are allowed to use pre-trained networks, including the one you mentioned. You are also allowed to fine-tune its weights. The things that the rule doesn’t allow is using models that are not free to use and publically available. I hope this helps. Let me know if you have any more questions.

Shubhamai

#4

Hi.

It seems that the missing faces are AI generated. At least some target faces seem to show real people. In case some photos show real people: Please tell if GDPR applies and if the people gave their explicit consent for using their biometric data.

Thank you. I am asking because I am based in the EU.

Best
pwiener

#5

Hi @pwiener,

All the images used are generated using a Stylegan3 model, which was trained on Flickr-Faces-HQ Dataset. As per the privacy policy of the Flickr-Faces-HQ Dataset, they were “careful to only include photos that - to the best of our knowledge - were intended for free use and redistribution by their respective authors. free for use and redistribution.”

In case, you find any images of real individuals, you can also let us know, we will be more than happy to remove the said images from our prepared datasets.

Hope that answers your question.

Best,
Mohanty