Sheep Facial Expression Pain Rating Scale: Using Convolutional Neural Networks

Analyzing facial expression of animals by convolutional neural network is the main focus of this research.Because manual evaluation of sheep facial expression is lack of accuracy,time-consuming,and monotonous.While pain level estimation from facial expression is an efficient and reliable mark of sheep life.Hence,the fundamental of deep learning is the recent advancement of convolutional neural network in computer vision which helps to classify facial expression fast and accurate.On the other hand,convolutional neural network effects by small samples,and it is due to Gaussian and Impulse noises in the input.In the methodology,the first stage is the elimination of combined Gaussian and impulse noises in digital image processing with preservation of image details and suppression of noise are challenging problem.For this purpose,a new filter which is median filters combined with convolutional neural network for Gaussian and salt & pepper noises.The previous methods are application dependents;some used for impulse noise and other employed only for Gaussian noise.The elimination of Gaussian and impulse noise completed into two steps first is the detection of impulse noise with the rejection of noise by employed of 3 × 3 and 5 × 5 window size median filters.In the second step,removal of Gaussian noise performed by residual learning denoising convolutional neural network.It is very favorable and the ability of learning and denoising performance in the field of digital image processing.Denoising convolutional neural network also has active Gaussian noise with an unknown level of noise.While in the second stage of proposed technique,monitoring of sheep life taken place with adequate assessment in natural habits is essential for management.In this research,we proposed a sheep face dataset and framework that uses transfer learning with fine-tuning for automating the classification of normal and abnormal sheep face images.The state-of-the-art convolutional neural networks based architectures are used to train sheep face dataset.We used data augmentation,L2 regularization,and fine-tuning to train the models.In the experimental work,the noise removal shown that the proposed method can achieve low loss and root mean square error during training,high peak signal to noise ratio,low mean square error,image quality assessment with good quality and mean absolute error for close prediction between denoised and original color images.Experimental work for sheep facial expression as a normal and abnormal face detection on sheep face dataset performed by using pre-trained models and,results achieved 100% training,99.69% validation and 100% testing accuracy by used VGG16 model.While employing other pre-trained models,we achieved 93.10% to 98.4% accuracy.In last,VGG16 and ResNet50 employed for five facial expression and got 100% and 85% accuracy respectively.Our proposed model is suitable and to test on a large dataset and can assist other animal life with higher accuracy,save time and expenses.