Zhuang Liu, Hanzi Mao, Chao-Yuan Wu
Christoph Feichtenhofer, Trevor Darrell, Saining Xie
2022 · arXiv
The "Roaring 20s" of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually "modernize" a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets.
Metadata: arXiv. Abstracts retain their original authors’ rights.
Original HN comments, with labels from simple text rules. A label describes the whole comment and may miss context or irony. Read the discussion to make up your own mind. About the labels
6 selected excerpts · 6 total mentions
Perhaps; in a lot of cases the architecture barely matters. Transformers took a lot of extra tricks to get working well; the ConvNext paper showed that applying those same tricks to convolutional networks can fully close the gap. https://arxiv.org/abs/2201.03545
> The rule in ML There's definitely attempts to revive things (in the general sense, not just GANs), but most successes appear to come from large labs dedicating equal computational resources to the older models and often by changing names. This can make things more confusing and make things appear to be changing faster, but once you can see this, you'll have an easier time keeping up (so being new, watch out for this). I'll give some examples that are easier to read[0-2] (i.e. don't need expert knowledge to understand the nuances). As an insider (ML researcher), my complaint isn't so much about that we have a large proportion of people chasing one specific avenue, it is…
that we gate keep newer methods. I think you can see a similar effect on HN when new models are proposed. They are trashed due to lack of beating existing models (this is true even beyond ML!). There will always be reasons to critique works and I don't want discourage criticism, but I do want to discourage dismissal. It hinders progress, because progress is made in small steps, not leaps and bounds. I think this can get confusing for someone entering the field (I'm sorry if I've misjudged, I'm inferring from the comments). > On the FFHQ point This is an excellent question that unfortunately I don't know the answer to. I think you'll find this work helpful[3], it has the largest human study. But despite its name, StyleNAT performed best in specifically FFHQ. What I would say is that there are good arguments to make the diffusion models are better at representing a diversity of images (making them well suited for things like art generation) but theoretically GANs are approximating your full density distribution. There's some talks by Goodfellow discussing this but I don't recall which ones. > I didnt know they wrote custom kernels As you've probably found, the StyleGAN code is not the easiest to read lol. Since you're using pytorch, you can find them here[4]. I encourage you to look at these, especially if you've never seen CUDA code before. Because the biggest takeaway will be that you'll see how easy it is to add a custom kernel, and given the earlier comment you'll see the utility ^__^ I'm not going to discourage you from trying triton, but I'll note that pytorch's compile goes a long way. Definitely _start_ there (see TensorRT). > What does it mean to have a backbone? Does it just mean the underlying architecture used in the method? Exactly! So in the case of a diffusion model it is the UNet (the neural network part, and specifically this part estimates the parameters for the probability distribution. If this doesn't make sense now, it will later. If you are struggling to understand diffusion models after spending some time reading the papers, come back to this comment). You'll also find the term "backbone" used in application based models such as in Semantic Segmentation, Object Detection, Pose Estimation, and much more. In those cases, these are typically pretrained, so recognize this as a hyper-parameter. > Also, on the decoder only vs encoder-decoder point I'm going to say something frustrating. In short: yes. If we get a but more nuanced: no. If we get really nuanced: yes. I know this isn't a great answer, but it can be really difficult to understand. On the surface, yes because you need to encode the variable and your model needs to transform a dimension starting at R^N and ending in R^N. While a (I have to stress, colloquial[5]) GAN transforms a R^M space to R^N where M << N. With more nuance you can argue it is the backbone. But to be detailed, you'll find that there are fundamental factors placing computational bounds on the theoretical performance of these architectures. To get there you'll need to carefully study Goodfellow's original paper (some follow-ups expanding on the analysis) as well as the "original" diffusion paper by Sohl-Dickstein[6] (quotes because this is debatable, but the claim has reasonable merit), and you should become familiar with Aapo Hyvärinen[7]. The last is by far the hardest part and the confusion is normal. I know quite a few well renowned and intelligent people who struggled (personally I went through the "this is hard", "this is easy", "ah fuck, I actually don't understand anything" cycle for a bit. But that's just a signal that you're progressing :). > Thanks for the detailed comment, you've given me a lot to think about. Great! I hope this can help provide direction (I read your other post). The best and worth thing about machine learning is that there's so much depth. It can both be intimidating and easy to miss. But if you're passionate about learning (as it appears) you'll find the knowledge gained is highly rewarding, if unfortunately hard to gain. And I apologize for being so verbose. It is a bad habit. [0] Diffusion Models Beat GANs on Image Synthesis (https://arxiv.org/abs/2105.05233) The two authors are rockstars. Given your blog post, I think you'll enjoy a lot of their work. Which includes diffusion. [1] ResNet strikes back: An improved training procedure in timm (https://arxiv.org/abs/2110.00476) Again, all three are great people to follow. You won't see many papers by Wightman, but you'll see his work with (now) Hugging Face. Notably he's one of the most important players in ViTs. [2] A ConvNet for the 2020s (https://arxiv.org/abs/2201.03545) and ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders (https://arxiv.org/abs/2301.00808) [3] Exposing flaws of generative model evaluation metrics and their unfair treatment of diffusion models (https://arxiv.org/abs/2306.04675) [4] https://github.com/NVlabs/stylegan2-ada-pytorch/tree/main/to... [5] I'm sorry, I still have difficulties explaining this, especially simply. There's a lot of points here. But one is easy to understand and what I mentioned before: GAN is a training method, not an architecture. A bit more nuance will be found by reading this far underappreciated work: https://arxiv.org/abs/1912.03263. The last point I want to mention is to never forget that "generative" is a general term and these models are good for generating __data__. Images are data, but to think this is the only type of data a GAN (or any model. I literally mean any[3]) can generate is naive. All of this gets harder to explain and I don't have the skill to do so in a simple manner and am afraid it will just come off as a rant. [6] Deep Unsupervised Learning using Nonequilibrium Thermodynamics (https://arxiv.org/abs/1503.03585) [7] "Score matching" and "Noise-contrastive estimation" will be the most beneficial https://www.cs.helsinki.fi/u/ahyvarin/papers/
Academic papers often include an introduction which summarises the context in which the paper is written. Consider https://arxiv.org/pdf/2201.03545v2.pdf where the paper says "For many decades, this has been the default use of ConvNets, generally on limited object categories such as digits [43], faces [58, 76] and pedestrians [19, 63]. Entering the 2010s, the region-based detectors [23, 24, 27, 57] further elevated ConvNets to the position of being the fundamental building block in a visual recognition system." This shows the authors are familiar with the field, avoids inadvertent plagiarism, allows them to make it clear precisely how their contribution contributes to…
knowledge, shows people from funding bodies that this is a cutting edge and important field of research, and points readers in the right direction if they want to see earlier ideas, or ideas in a more applied context. If among those 92 citations should be a few papers from your boss and your colleagues, nobody will see that as unusual - as long as they're at least marginally relevant.
Researchers love the ViT and all its varieties, but a reminder for those following at home that ConvNets still work and scale fine, depending on your requirements. For example, in "A ConvNet for the 2020s"[1], the authors are able to scale up ConvNets to the sizes of ViTs. From the abstract: "Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation" [1] https://arxiv.org/abs/2201.03545
> ViT models are outperforming CNNs in terms of computational efficiency and accuracy, achieving highly competitive performance in tasks like image classification, object detection, and semantic image segmentation. Since then, this has been show to be untrue. Using more modern training techniques along with depthwise convolutions (https://arxiv.org/abs/2201.03545) results in equal if not better performance on vision tasks. Improved training methodologies have also been shown to boost the accuracy of ResNet50 - an 6-year-old pure convolutional architecture - on ImageNet-1k by over 5% (https://arxiv.org/abs/2110.00476). Pure ViTs are also more difficult to train when compared with traditional convnets, although this has since then been somewhat remedied by Swin (https://arxiv.org/abs/2103.14030).
The performance advantages of transformer architectures have little to do with "transformers" themselves, but rather other architecture changes that were made along the way. They can be matched or beat in performance by tweaked conventional CNNs: https://arxiv.org/abs/2201.03545
More from the same topics.
Sébastien Bubeck, Varun Chandrasekaran, Ronen Eldan +11
Artificial intelligence (AI) researchers have been developing and refining large language models (LLMs) that exhibit remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition.
Tom B. Brown, Benjamin Mann, Nick Ryder +28
Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task.
Chiyuan Zhang, Samy Bengio, Moritz Hardt +2
Despite their massive size, successful deep artificial neural networks can exhibit a remarkably small difference between training and test performance.
A. Pluchino. A. E. Biondo, A. Rapisarda
The largely dominant meritocratic paradigm of highly competitive Western cultures is rooted on the belief that success is due mainly, if not exclusively, to personal qualities such as talent, intelligence, skills, efforts or risk taking.