TO-DO
The dataset was used in this project contains more than 50000 movie reviews, and it split up into Train, Validation, and Test sets already. All the movie reviews are long sentences (most of them are longer than 200 words). Also, each review was labeled as 1 (positive review) or 0 (negative review). The dataset was originally introduced in [1], but its .csv file can be downloaded from here.
TO-DO
- Andrew L. Maas, Raymond E. Daly, Peter T. Pham, Dan Huang, Andrew Y. Ng, and Christopher Potts. (2011). Learning Word Vectors for Sentiment Analysis. The 49th Annual Meeting of the Association for Computational Linguistics (ACL 2011).
TO-DO
TO-DO
TO-DO
TO-DO
An offline signature verifier model should be trained using the Convolutional Neural Networks (CNNs). The problem can be interpreted in many ways, such as classifying signatures based on their signatories or detecting that a signature is genuine or forgery. But, the first interpretation was considered at the implementation.
The UTSig [1] dataset was used in this project, and it can be downloaded from here. UTSig has 115 classes containing: 27 genuine signatures, 3 opposite-hand signed samples, and 42 simple forgeries. Each class belongs to one specific authentic person.
The Inception V3 was used as a feature extractor in this project. It's trained by Google on more than a million images from the ImageNet database to classify images into 1000 object categories.
- Amir Soleimani and Kazim Fouladi and Babak Nadjar Araabi (2016). UTSig: A Persian Offline Signature Dataset. https://arxiv.org/abs/1603.03235