Kayıkçı, Şafak2021-06-232021-06-232018978-1-5386-7893-0https://hdl.handle.net/20.500.12491/9688https://ieeexplore.ieee.org/document/85663183rd International Conference on Computer Science and Engineering (UBMK) -- SEP 20-23, 2018 -- Sarajevo, BOSNIA & HERCEGCAPTCHAs arc used for separating humans and machines. They arc tests which can be solved by people but not by computers. With deep learning methods, this tests can be passed automatically. The success of Convolutional Neural Networks is related to the dataset size and quality. However, it is difficult to conclude with an inadequate training data. In this study, a convolutional neural network implementation is proposed for resolving CAPTCHAs without any human interaction with images derived from plugin. An active deep learning approach that is capable of retrain the network and present consequences on an auto-generated captcha dataset proposed. Outcomes show that the performance of the network may be significantly improved with the efficaciously classified test sets.eninfo:eu-repo/semantics/closedAccessArtificial IntelligenceCaptchaDeep Neural NetworkImage ProcessingConference Object41442-s2.0-85060610535N/AWOS:000459847400009N/A