A special kind of communication that helps close the gap between the general public and those with hearing difficulties is sign language. It is crucial for many communities since it makes it possible for those who are hard of hearing to communicate. There are many different signs used in sign languages, and they are all distinguished by variations in the hand forms, hand position, gestures, face expression, and body part utilized to take certain meaning. A major obstacle in the field of computer vision research is the intricacy of visual sign language detection. This paper offers an automatic and accurate recognition method for Arabic Sign Language characters using k-nearest neighbor (KNN) and multiple transfer learning models. The study’s dataset includes 54,048 letter-themed photos. The study’s findings show that InceptionV3 fared better than previous pretrained models, obtaining an astounding 99.4% accuracy value and a 0.6 loss value without the need for overfitting. The remarkable output metrics demonstrate how well InceptionV3 distinguishes itself in Arabic character recognition and emphasize how resilient it is to overfitting. This makes it more promising for use in subsequent studies on the recognition of Arabic Sign Language.