Google Cloud Vision is readily used for major purposes such as label detection face recognition mood analysis, object detection content filtering and that is to a certain extent. The efficiency of any system is based on the fact that how well the system is performing in suboptimal conditions in case of Google Cloud Vision the suboptimal working condition include the use of noisy images instead of perfect ones. This paper deals with how this Google Cloud Vision works under noisy images and how robust the system stays under these conditions. This API generates different outputs by adding different noises with different intensity in noise. It is clearly observed that with the mean value of 20% impulse noise and 0.1 variance Gaussian noise, the API can be easily misguided in predicting the actual label and text for the images. A better and accurate outcome can be obtained by pre-processing and validating the image for any noise and denoising an image up to some extent for a better and accurate outcome which could be more beneficial than updating the currently working algorithm.

Google Cloud Vision, Robustness, Noisy Images, Gaussian Noise, Impulse Noise

[1] F. R¨ohrbein, P. Goddard, M. Schneider, G. James, and K. Guo, “How does image noise affect actual and predicted human gaze allocation in assessing image quality?” Vision research, vol. 112, pp. 11–25, 2015.
[2] K. Simonyan and A. Zisserman. “Very deep convolutional networks for large-scale image recognition”. In Proceedings of conference paper at ICL, 2015
[3] I. Vasiljevic, A. Chakrabarti, and G. Shakhnarovich, “Examining the impact of blur on recognition by convolutional networks,” arXiv preprint arXiv:1611.05760, 2016.
[4] S. Diamond, V. Sitzmann, S. Boyd, G. Wetzstein, and F. Heide, “Dirty pixels: Optimizing image classification architectures for raw sensor data,” arXiv preprint arXiv:1701.06487, 2017.
[5] Papernot, N., M Cdaniel, P., Jha, S., Fredrikson, M., Celik, Z. B.,and Swami, A. “The limitations of deep learning in adversarial settings”. In Proceedings of the 1st IEEE European Symposium on Security and Privacy”, arXiv preprint Xiv:1511.07528, 2016.
[6] Ratnesh Kumar Shukla, Ajay Agarwal, Anil Kumar Malviya, "An Introduction of Face Recognition and Face Detection for Blurred and Noisy Images", International Journal of Scientific Research in Computer Science and Engineering, Vol.6, Issue.3, pp.39-43, 2018
[7] Dougherty G. “Digital Image Processing for Medical Applications,” second ed., Cambridge university press, 2010.
[8] Boyat, A. and Joshi, B. K. “Image Denoising using Wavelet Transform and Median Filtering”, 2013 Nirma University International Conference on Engineering (NUiCONE), Ahmedabad, pp. 1-6, 2013.doi: 10.1109/NUiCONE.2013.6780128
[9] Mandeep Kaur, Balkrishan Jindal, "Improved Sparse matrix Denoising Techniques using affinity matrix for Geographical Images", International Journal of Scientific Research in Computer Science and Engineering, Vol.5, Issue.5, pp.51-56, 2017
[10] Priyanka Kamboj, Versha Rani,” Brief study of various noise model and filtering techniques”, Journal of Global Research in Computer Science, vol.4, No.4, pp.166-171, April 2013.
[11] Monika Raghav, and Sahil Raheja,” Image Denoising Techniques: Literature Review”, International Journal of Engineering and Computer Science, vol.3, pp. 5637-5641, Issue 5, May 2014.
[12] Joshi, A., Boyat, A. and Joshi, B. K. “Impact of Wavelet Transform and Median Filtering on removal of Salt and Pepper noise in Digital Images,” IEEE International Conference on Issues and Challenges in Intelligant Computing Teachniques, Gaziabad, India, 2014