Summary
In this chapter, we saw the power of CNNs. We began by examining the challenges faced by DNNs for visual recognition tasks. Next, we journeyed through the anatomy of CNNs, zooming in on the various moving parts, such as the convolutional, pooling, and fully connected layers. Here, we saw the impact and effect of different hyperparameters, and we also discussed the boundary effect. Next, we moved on to using all we learned to build a real-world weather classifier using two DNNs and a CNN. Our CNN model outperformed the DNNs, showcasing the strength of CNNs in handling image-based problems. Also, we discussed and applied some TensorFlow functions that streamline data preprocessing and modeling when we are working with image data.
By now you should have a good understanding of the structure and operations of CNNs and how to use them to solve real-world image classification problems, as well as utilizing various tools in TensorFlow to effectively and efficiently preprocess image...
