In recent days image processing people apply image processing to many problems for effective reasons. So, this paper describes the work on defect identification of products which are manufactured in industries using image processing as backbone. In the existing methods, industries use manual checking by different gauges to identify the defects in products. In this proposed model, image processing method is employed to identify various defects in products. First the images are captured using an image sensor for training the dataset. The captured images are classified with different classes and trained using Recurrent Neural Networks (RNN). While real time capturing, first the picture of the product is captured and compared with the existing trained dataset. For testing, gear is employed as a subject material. Here we have classified different cases into three as tooth missing, scratched gear and perfect gear. Python is the base language for the process which takes place for building the dataset and entire flow of the technique. The image processing algorithm is built over TensorFlow to identify the defect with the support of the dataset.