Blog

The latest from Google Research

Revisiting the Unreasonable Effectiveness of Data

Tuesday, July 11, 2017

Posted by Abhinav Gupta, Faculty Advisor, Machine Perceptiondeep learningAlexNetResNetImageNet

While GPU computation power and model sizes have continued to increase over the last five years, the size of the largest training dataset has surprisingly remained constant.

Revisiting Unreasonable Effectiveness of Data in Deep Learning Era

Better Representation Learning Helps. Our first observation is that large-scale data helps in representation learning which in-turn improves the performance on each vision task we study. Our findings suggest that a collective effort to build a large-scale dataset for visual pretraining is important. It also suggests a bright future for unsupervised and semi-supervised representation learning approaches. It seems the scale of data continues to overpower noise in the label space.

Performance increases logarithmically based on volume of training data. We find there is a logarithmic relationship between performance on vision tasks and the amount of training data used for representation learning.

Object detection performance when pre-trained on different subsets of JFT-300M from scratch. x-axis is the dataset size in log-scale, y-axis is the detection performance in mAP@[.5,.95] on COCO-minival subset.

Capacity is Crucial. We also observe that to fully exploit 300M images, one needs higher capacity (deeper) models. For example, in case of ResNet-50 the gain on COCO object detection benchmark is much smaller (1.87%) compared to (3%) when using ResNet-152.

New state of the art results. Our paper presents new state-of-the-art results on several benchmarks using the models learned from JFT-300M. For example, a single model (without any bells and whistles) can now achieve 37.4 AP as compared to 34.3 AP on the COCO detection benchmark.