2019
Knowledge Distillation and Incremental Learning
Published:
A detailed introduction to Knowledge distillation from complex neural networks and a much needed end to end Incremental Learning technique.
Adapting to Cross-Domain Data in Object Detection: OMNIA Faster R-CNN
Published:
This is a brief review of the paper titled “OMNIA Faster R-CNN: Detection in the wild through dataset merging and soft distillation”. This is a fairly recent paper and solves a problem which is quite commonplace with object etectors.
The New Microsoft Terminal
Published:
So, the Microsoft Store now finally has the Preview version of the much hyped Windows Terminal, and having used it for some time now, I have come to like this single-window multi-tabbed terminal. I have been using ConEmu for some time now, and although it also has the multi-tab feature, it kind of gets slow at times. The Windows Terminal feels really snappy and uses lower memory overall. The Terminal also seems to have a better UI, but I feels it lacks the feature of glass-slab like transparency. Windows Terminal has the acrylic transparency feature which blotches out the background for most part and this is a feature that I strongly feel should be incorporated in the final version.
Python Memory Management
Published:
Python uses a specially tuned memory allocator on top of a general purpose allocator. This suits the universal object model in Python really well, while setting up a nice hierarchy in the memory management system, beginning from the OS’ Virtual Memory Manager to Python’s own object allocator. In this post, I have tried to build up a few concepts while leading up to detailed discussion about the special Python memory allocator. Note: I do not discuss the OS VMM or C malloc.
Test Post
Published:
Test.