Top all-time · #7594
Dmitry Ulyanov
Personal blog · dmitryulyanov.github.ioaiprogramming
Dmitry Ulyanov's personal blog at dmitryulyanov.github.io covers ai and programming. Since 2016, 10 different Hacker News readers have linked it 10 times, most often “Deep image prior 'learns' on just one image”. The last post was in 2017, but the archive is still widely read.
Too few sampled comments (4) to say how often HN recommends it.?
Why HN links it
Commenters cite the deep_image_prior project page, which shows an untrained convolutional network can handle image denoising and inpainting from a single image. It is used as supporting evidence in discussions of priors and compression.
Summarised by Claude from a sample of 4 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Technical blog
Most-linked posts?
- Deep image prior 'learns' on just one image7 people · 7 mentions · first linked 2017-11-30
- Neural Point-Based Graphics2 people · 2 mentions · first linked 2019-07-24
What Hacker News says?
Somewhat tangentially, some recent work showed that a lot of problems with images (e.g. denoising, upsampling, inpainting, etc...) can be solved very efficiently with no training set at all: https://dmitryulyanov.github.io/deep_image_prior This work shows that the network architecture is a strong enough prior to effectively learn this set of tasks from a single image. Note that there is no pretraining here whatsoeve…
Comment linking /deep_image_prior
on “Autopsy of a deep learning paper”Submitted Deep image prior 'learns' on just one image as a story · 782 points
HN discussionThere was another submission recently that kind of does what you described. Click on the visualizations to see the iterations: https://dmitryulyanov.github.io/deep_image_prior . The ELI5 is that prior knowledge of what images look like can be used to reconstruct images w/o corruption. For non-images, this idea is pretty old and sounds a lot like compression algorithms that share a pre-defined dictionary. For example…
Comment linking /deep_image_prior
on “DeepZip: Lossless Compression Using Recurrent Networks [pdf]”Submitted Audio texture synthesis and style transfer as a story · 85 points
HN discussion> You can start with a random neural net. It's not exactly empty, but it's ok. One interesting thing about random NNs is how much they can already do. For example, you can do single-shot image inpainting or superresolution with an untrained randomly-wired CNN: https://dmitryulyanov.github.io/deep_image_prior In RL, you can use randomly sampled NNs to create various artificial arbitrary 'reward functions' to force an…
Comment linking /deep_image_prior
on “Reinforcement learning’s foundational flaw”Submitted Neural Point-Based Graphics as a story · 5 points
HN discussionWell likely, but this title is very strong for whatever arguments that is given. The better question to ask would what is possible with neural network in next 10 years. I don't know if anyone could have predicted neural network could pass turing test even 5 years before. Now GPT 3 could talk for 100s of dialogues coherently. > If a baby ibex can clamber down the side of a mountain shortly after birth, why shouldn’t…
Comment linking /deep_image_prior
on “Deep Learning Alone Isn’t Getting Us to Human-Like AI”Submitted Neural Point-Based Graphics as a story · 4 points
HN discussionSubmitted Deep Image Prior as a story · 3 points
HN discussionSubmitted Deep Image Prior (2017) as a story · 2 points
HN discussion
