Computer Vision Modeling on the Cloud

Training state-of-the-art convolutional neural networks can require significant computer resources. It may take several hours or days to train a network depending on the number of training samples, network configuration, and available hardware resources. A single GPU may not be feasible to train a complex network involving large numbers of training images. The models need to be trained on multiple GPUs. Only a limited number of GPUs can be installed on a single machine. A single machine with multiple GPUs may not be sufficient for training on a large number of images. It will be faster if the model is trained on multiple machines with each machine having multiple GPUs.