Summary of DeepLabv3 paper

The authors revisited atrous convolution, a powerful tool to adjust filter's field-of-view to incorporate multi-scale context and control the resolution of feature responses computed of Deep Convolutional Neural Networks(DCNNs).

Summary of BatchNorm paper

The authors identified internal Covariance Shift problem and addressed this issue by normalizing layer inputs and presented novel method called BatchNorm for dramatically accelerating the training of deep networks.

Summary of FCN paper

This paper introduced the idea of converting classification networks into fully convolutional networks that produce coarse outputs. Then these coarse outputs are connected to dense pixels for pixelwise prediction. The...

Semantic Image Search

In the previous post, I described about APIs, HTTP requests and responses using curl. Whereas, in this post we develop an interesting project based on cool web APIs available. The...

Beginner's guide to APIs and curl

If you’re in tech world, you should have heard about web APIs by now. You might be a newbie like me and still figuring out what exactly APIs are. I...