ImageNet
A large-scale visual dataset constructed by Fei-Fei Li's team at Stanford University, containing over 14 million annotated images across 1,000 categories. The dataset catalyzed the 2012 AlexNet breakthrough and serves as a landmark benchmark in deep learning-based computer vision research.
Core Facts
In 2012, AlexNet won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) by a margin of nearly 11 percentage points over the second-place entry, reducing the Top-5 error rate from 26% to 15.3%
Deep learning achieved a breakthrough performance in the 2012 ImageNet competition, driving the revival of connectionism
The transfer learning strategy uses ImageNet pre-trained weights as a foundation and fine-tunes on agricultural-specific datasets, enabling high recognition accuracy to be achieved with relatively little training data.
Timeline (last 90 days)
Guo et al. in 2017 found that deep models on ImageNet had average confidence far higher than actual accuracy, and temperature scaling is a simple and effective post-processing calibration correction method.
MiniDog's pre-training uses 26,000 dog images from ImageNet, and generates text descriptions for them.
The paper tested GCD on CIFAR-10 and ImageNet, covering three architectures, ResNet-18, ResNet-50, and ViT-B/16, and evaluated four mainstream attack methods.
The Context Encoder's encoder is trained from scratch with random weights, trained solely on the inpainting task rather than for ImageNet classification.
The V1 approach uses ImageNet-pretrained EfficientNet-B2 to generate embeddings of font-rendered images and retrieves matches via nearest-neighbor search.
Early benchmarks such as ImageNet and GLUE reshaped the optimization direction of the entire machine learning research field.
A handful of datasets such as ImageNet, nuScenes, KITTI, and Waymo Open Dataset have long dominated evaluation leaderboards for autonomous driving perception, detection, and planning tasks.
EfficientNet-B3 has an input resolution of 300×300 and achieves a Top-1 accuracy of approximately 81.6% on ImageNet.
Almost all mainstream visual backbone networks (ResNet, EfficientNet, ViT, etc.) have been pretrained on ImageNet.
ImageNet is a classic dataset for image classification pretraining, containing over one million images.
27 more timeline events
All Facts (20)
In 2012, AlexNet won the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) by a margin of nearly 11 percentage points over the second-place entry, reducing the Top-5 error rate from 26% to 15.3%
90%VerifiedDeep learning achieved a breakthrough performance in the 2012 ImageNet competition, driving the revival of connectionism
85%VerifiedThe transfer learning strategy uses ImageNet pre-trained weights as a foundation and fine-tunes on agricultural-specific datasets, enabling high recognition accuracy to be achieved with relatively little training data.
80%VerifiedAlexNet在ImageNet竞赛中夺冠正式开启了深度学习时代
80%VerifiedIn 2012, Hinton's team's AlexNet won the ImageNet competition by an overwhelming margin.
80%UnverifiedThe choice of pretraining (ImageNet-1K vs ImageNet-21K, supervised vs self-supervised like MAE/DINO) significantly impacts downstream performance, potentially causing several percentage points of difference
80%UnverifiedThe construction of ImageNet relied on nearly 50,000 annotators from 167 countries on the Amazon Mechanical Turk crowdsourcing platform, at a total cost of approximately $22 million.
60%UnverifiedGuo et al. in 2017 found that deep models on ImageNet had average confidence far higher than actual accuracy, and temperature scaling is a simple and effective post-processing calibration correction method.
50%UnverifiedMiniDog's pre-training uses 26,000 dog images from ImageNet, and generates text descriptions for them.
50%UnverifiedThe paper tested GCD on CIFAR-10 and ImageNet, covering three architectures, ResNet-18, ResNet-50, and ViT-B/16, and evaluated four mainstream attack methods.
50%UnverifiedThe Context Encoder's encoder is trained from scratch with random weights, trained solely on the inpainting task rather than for ImageNet classification.
50%UnverifiedThe V1 approach uses ImageNet-pretrained EfficientNet-B2 to generate embeddings of font-rendered images and retrieves matches via nearest-neighbor search.
50%UnverifiedEarly benchmarks such as ImageNet and GLUE reshaped the optimization direction of the entire machine learning research field.
50%UnverifiedA handful of datasets such as ImageNet, nuScenes, KITTI, and Waymo Open Dataset have long dominated evaluation leaderboards for autonomous driving perception, detection, and planning tasks.
50%UnverifiedEfficientNet-B3 has an input resolution of 300×300 and achieves a Top-1 accuracy of approximately 81.6% on ImageNet.
50%UnverifiedImageNet is a classic dataset for image classification pretraining, containing over one million images.
50%UnverifiedAlmost all mainstream visual backbone networks (ResNet, EfficientNet, ViT, etc.) have been pretrained on ImageNet.
50%UnverifiedBenchmark saturation phenomenon: after ImageNet was conquered, ObjectNet and ImageNet-A emerged; after GLUE was saturated, SuperGLUE appeared, followed by BIG-Bench and BIG-Bench Hard.
50%UnverifiedGridSFM trains a single model on more than 150 base grid topologies and approximately 500,000 scenarios, in a training approach similar to ImageNet pretraining.
50%UnverifiedFei-Fei Li was previously the director of Stanford University's AI lab and is the primary driving force behind the ImageNet dataset.
50%