Open Access

Fig. 7

image

Download original image

Process of labeling or pseudo-labeling for unlabeled samples. Following the training phase in the ITS, we select K hard samples from the pool of (N−testM) unlabeled data and label them with expert annotations. Subsequently, after the training phase in the ALS, we choose V over-threshold samples from (N−testM − K) unlabeled data to assign pseudo-labels. Moving forward, after the training process in SSLS in the r-th iteration, we again select V samples from the pool of (N−testMK) unlabeled data that exceed the predetermined threshold and assign pseudo labels. In the (r + 1)-th iteration, we retrain the semi-supervised training model from scratch, refining the performance of our models in subsequent iterations.

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.