Skip to content
The methodOpen data
Back to the collection

YOLOv4: Optimal Speed and Accuracy of Object Detection

Alexey Bochkovskiy, Chien-Yao Wang, Hong-Yuan Mark Liao

2020 · arXiv

Computer visionMachine learning
Read the paper
HN score
0
Mentions
4
Readers
3
Months
4
What’s in a score?

From the abstract

There are a huge number of features which are said to improve Convolutional Neural Network (CNN) accuracy. Practical testing of combinations of such features on large datasets, and theoretical justification of the result, is required. Some features operate on certain models exclusively and for certain problems exclusively, or only for small-scale datasets; while some features, such as batch-normalization and residual-connections, are applicable to the majority of models, tasks, and datasets. We assume that such universal features include Weighted-Residual-Connections (WRC), Cross-Stage-Partial-connections (CSP), Cross mini-Batch Normalization (CmBN), Self-adversarial-training (SAT) and Mish-activation. We use new features: WRC, CSP, CmBN, SAT, Mish activation, Mosaic data augmentation, CmBN, DropBlock regularization, and CIoU loss, and combine some of them to achieve state-of-the-art results: 43.5% AP (65.7% AP50) for the MS COCO dataset at a realtime speed of ~65 FPS on Tesla V100. Source code is at https://github.com/AlexeyAB/darknet

Metadata: arXiv. Abstracts retain their original authors’ rights.

In the conversation

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

4 selected excerpts · 4 total mentions

edouard-harris
Neutral mention · rule
That Tim Urban piece is great. It's also an interesting time capsule in terms of which AI problems were and were not considered hard in 2015 (when the post was written). From the post: > Build a computer that can multiply two ten-digit numbers in a split second—incredibly easy. Build one that can look at a dog and answer whether it’s a dog or a cat—spectacularly difficult. Make AI that can beat any human in chess? Done. Make one that can read a paragraph from a six-year-old’s picture book and not just recognize the words but understand the meaning of them? Google is currently spending billions of dollars trying to do it. Hard things—like calculus, financial market…
Read the full comment
strategy, and language translation—are mind-numbingly easy for a computer, while easy things—like vision, motion, movement, and perception—are insanely hard for it. The children's picture book problem is solved; those billions of dollars were well-spent after all. (See, e.g., DeepMind's recent Flamingo model [1].) We can do whatever we want in vision, more or less [2]. Motion and movement might be the least developed area, but it's still made major progress; we have robotic parkour [3] and physical Rubik's cube solvers [4], and we can tell a robot to follow simple domestic instructions [5]. And Perceiver (again from DeepMind [6]) took a big chunk out of the perception problem. Getting a computer to carry on a conversation [7], let alone draw art on par with human professionals [8], weren't even mentioned as examples, so laughably out of reach they seemed in the heathen dark ages of... 2015. And as for recognizing a cat or a dog — that's a problem so trivial today that it isn't even worth using as the very first example in an introductory AI course. [9] If someone re-wrote this post today, I wonder what sorts of things would go into the "hard for a computer" bucket? And how many of those would be left standing in 2029? [1] https://arxiv.org/abs/2204.14198 [2] https://arxiv.org/abs/2004.10934 [3] https://www.youtube.com/watch?v=tF4DML7FIWk [4] https://openai.com/blog/solving-rubiks-cube/ [5] https://say-can.github.io/ [6] https://www.deepmind.com/open-source/perceiver-io [7] https://arxiv.org/abs/2201.08239v2 [8] https://openai.com/dall-e-2/ [9] https://www.fast.ai/
Original comment DeepMind: A Generalist Agent
GhettoComputers
Neutral mention · rule
You’ll want a breakout board to connect it to motors, or see other projects. It’s possible to make a separate mobile platform for the sensors on the phone and have it gather info. You’re going to want to look for NPUs and probably not GPUs. If you don’t have any idea check this out for inspiration. https://arxiv.org/abs/2004.10934 The DJI founders wanted engineers to be seen as hero’s instead of sports stars. He sponsors engineers to complete here, the prize is their dream job. https://www.robomaster.com/en-US
Original comment Pine64 and Radxa's new Pi CM4-compatible boards
rocauc
Neutral mention · rule
EfficientDet was open sourced March 18 [1], YOLOv4 came out April 23 [2], and now YOLOv5 is out only 48 days later. In our initial look, YOLOv5 is 180% faster, 88% smaller, similarly accurate, and easier to use (native to PyTorch rather thank Darknet) than YOLOv4. [1] https://venturebeat.com/2020/03/18/google-ai-open-sources-ef... [2] https://arxiv.org/abs/2004.10934
Original comment YOLOv5: State-of-the-art object detection at 140 FPS
rocauc
Neutral mention · rule
YOLOv4 was published[1] on April 23 with COCO weights, but there haven't yet been resources on how to adapt its architecture to your own domain. This post walks through setting up Darknet and training on a custom dataset in Colab. FWIW, in our tests, we saw the highest mAP (89.5) on a new task[2] compared to EfficientDet and YOLOv3. [1] https://arxiv.org/abs/2004.10934 [2] https://public.roboflow.ai/object-detection/bccd
Original comment Training YOLOv4 on a Custom Dataset

Keep following the thread.

More in Computer vision

More from the same topics.