ImageNet classification with deep convolutional neural networks
Problem
Accurately classifying 1.2 million high-resolution images into 1000 categories, a task previous methods handled poorly.
Why it matters
Reliable large-scale image recognition underpins many practical applications and was a long-standing AI challenge.
What the AI did
A large deep convolutional neural network with 60 million parameters was trained on GPUs, using dropout regularization to reduce overfitting.
Results
Top-1 error 37.5% and top-5 error 17.0% on LSVRC-2010 test data; winning top-5 error 15.3% in ILSVRC-2012 vs 26.2% for runner-up.
Impact
Beat the previous state of the art and won the ILSVRC-2012 competition.
Record
| Venue | Communications of the ACM |
|---|---|
| Published | 2017 |
| Work type | article |
| Authors | Alex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton |
| Citations | 106,345 (OpenAlex, at import time) |
| References | 27 |
| Model or method | deep convolutional neural network (five convolutional layers, three fully connected; not named in source) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1145/3065386 |
| Open access | bronze |
| Keywords | deep learning, convolutional neural networks, ImageNet, image classification, GPU, dropout |
Importance rating
major — millions of lives, an industry or a field
Large-scale image recognition on 1.2M images/1000 classes; beat prior state of the art and won ILSVRC-2012, a foundational AI challenge.
More in Software & Coding
| # | Paper | Year | Citations | Importance |
|---|---|---|---|---|
| 1 | Deep Residual Learning for Image Recognition | 2016 | 221,134 |
8
|
| 2 | Faster sorting algorithms discovered using deep reinforcement learning | 2023 | 175 |
8
|
| 3 | A deep learning‐based approach for software vulnerability detection using code metrics | 2022 | 42 |
7
|
| 4 | Deep Learning-Based Attack Detection and Classification in Android Devices | 2023 | 41 |
7
|
| 5 | Dota 2 with Large Scale Deep Reinforcement Learning | 2019 | 969 |
6
|