Systematic evaluation of CNN advances on the ImageNet
This work provides incremental insights for researchers and practitioners in computer vision by benchmarking CNN components to guide efficient network design.
The paper systematically evaluates the impact of various CNN architectural and learning modifications on ImageNet object categorization, finding that individual gains are largely independent and that 128x128 pixel images suffice for qualitative conclusions, speeding up results by an order of magnitude compared to standard 224 pixel images.
The paper systematically studies the impact of a range of recent advances in CNN architectures and learning methods on the object categorization (ILSVRC) problem. The evalution tests the influence of the following choices of the architecture: non-linearity (ReLU, ELU, maxout, compatibility with batch normalization), pooling variants (stochastic, max, average, mixed), network width, classifier design (convolutional, fully-connected, SPP), image pre-processing, and of learning parameters: learning rate, batch size, cleanliness of the data, etc. The performance gains of the proposed modifications are first tested individually and then in combination. The sum of individual gains is bigger than the observed improvement when all modifications are introduced, but the "deficit" is small suggesting independence of their benefits. We show that the use of 128x128 pixel images is sufficient to make qualitative conclusions about optimal network structure that hold for the full size Caffe and VGG nets. The results are obtained an order of magnitude faster than with the standard 224 pixel images.