Jekyll2025-08-08T04:10:38-04:00https://snknitin.github.io/feed.xmlPortfoliopersonal descriptionNitin Kishore Sai Samalasnk.nitin@gmail.comGreen Screen Image Composition Transfer2022-10-14T00:00:00-04:002022-10-14T00:00:00-04:00https://snknitin.github.io/projects/2022/10/project-5Using StableDiffusion for text2img background generation and MODNET/U2net model to seamlessly superimpose the isolated foreground using Alpha-matting and apply Nvidia’s FastPhotoStyle for an image stylization and smoothening
congratulations to @Nitin_wysiwyg, bka NikeZoldyck, whose 🥞FSDL course project on Green Screen Composition is one of 🤗 @huggingface's "Spaces of the Week"!
Built my first @huggingface space app as part of the @full_stack_dl course project. It allows you to do 2 things : 1) Remove bg and superimpose on another image using alpha matting 2) Image Style transfer using @nvidia's FastPhotoStyle
]]>Nitin Kishore Sai Samalasnk.nitin@gmail.comDeepfakeCapsuleGAN2018-10-08T00:00:00-04:002018-10-08T00:00:00-04:00https://snknitin.github.io/projects/2018/10/project-4Using GANS to generate images that are then used for deep fakes. Using Capsule networks instead of regular CNNs
]]>Nitin Kishore Sai Samalasnk.nitin@gmail.comImproving Open Domain Dialogue-Systems (Chatbots)2018-04-08T00:00:00-04:002018-04-08T00:00:00-04:00https://snknitin.github.io/projects/2018/04/project-3Built a seq2seq neural conversational model in PyTorch using attention with intention and a diversity promoting objective function to prevent irrelevant generic outputs’
]]>Nitin Kishore Sai Samalasnk.nitin@gmail.comMinimally-Constrained Multilingual Embeddings via Artificial Code-Switching2017-09-01T00:00:00-04:002017-09-01T00:00:00-04:00https://snknitin.github.io/projects/2017/09/project-2Created language agnostic word embeddings via artificial code-switching to share structure across languages for any NLP task when you have less labeled data.
]]>Nitin Kishore Sai Samalasnk.nitin@gmail.comDSR Reinforcement Learning to navigate a Labyrinth2016-12-24T00:00:00-05:002016-12-24T00:00:00-05:00https://snknitin.github.io/projects/2016/12/project-1Generalized Successor Representations from Neuroscience within an end-to-end deep reinforcement learning framework, comparing its efficacy to DQN on two diverse environments (Mazebase and DOOM) given raw pixel observations.’