Pattern Recognition based Hand Gesture Recognition model Using Faster R-CNN Inception V2 Model

Download Article
Thitupathi Jangapally , Dr. Tryambak Hiwarkar

Abstract- The real-time hand movement acknowledgment under unconstrained conditions is a problematic PC vision issue. The adjustment in light and non-uniform foundation condition makes it hard to perform ongoing hand motion acknowledgment activities. This paper shows a locale-based convolutional neural system for continuous hand signal acknowledgment. The custom dataset is caught under unconstrained situations. The Faster locale-based convolutional neural network (Faster-RCNN) with Inception V2 engineering is utilized to remove the highlights from the proposed area. The standard accuracy, standard review, and F1-score are broke down via preparing the model with a learning pace of 0.0002 for Adaptive Moment Estimation (ADAM) and Momentum analyzer, 0.004 for RMSprop streamlining agent. The ADAM optimization calculation brought about better accuracy, review, and F1-score esteems after assessing custom test information. For the ADAM analyzer with crossing point over association (IoU) =0.5:0.95, the watched standard exactness is 0.794, the standard review is 0.833, and the F1-score is 0.813. For an IoU of 0.5, the ADAM analyzer brought about 0.991 standard accuracies with an expectation season of 137ms.


Inception-V2, Hand gesture recognition, Convolutional Neural Network, Region proposal, Faster-RCNN