Confidence Estimation for Object Detection in Document Images

•Deep neural networks always require more labelled data to be trained.•Annotating is time-consuming, learning on a limited amount of data becomes necessary.•The data to annotate must be correctly chosen to optimize performance.•Objects confidence estimated using Monte Carlo dropout improves the detection.•Prediction confidence based on object statistics estimates the mean Average Precision.•Confidence estimators give better selections of the samples and lead to better models.