Bishwash Khanal's Projects
Implementation of real world application of GANs.
State-of-the-art papers for depth estimation of 360 images.
A curated list on piecewise-planar reconstruction
Reports and codes related to Bachelors Thesis.
Implementation of basic image processing and classical computer vision algorithms.
An implementation of BERT from scratch for finetuning on QA tasks.
Get to know more about me and my projects.
The public CGAL repository, see the README below
This repository is created to test my knowledge on ChatGPT and Stable Diffusion.
Implementation of CLIP from OpenAI using pretrained Image and Text Encoders.
Data visualization using Python
Meta's Database Engineer Capstone
Implementation of different state-of-the-art deep learning model from scratch.
The notebooks used as a part of several courses taught about LLMs on DeepLearning.ai.
Implementation of Detection Transformer (DETR) from scratch for object detection.
Chat with your document using state-of-the-art RAG practices, powered by LLMs.
A lightweight, easy-to-use, and efficient C++ library for processing and rendering 3D data
Things happening at Machine Learning table.
Attempt at Google's competitiom on Isolated Americal Sign Language Recognition through Fingerspelling hosted on Kaggle.
Road towards diffusion models.
Attempt at Google's competitiom on Isolated Sign Language Recognition hosted on Kaggle.
Implementation of Generative Pre-Training (GPT)-2 model from scratch.
Scripts for fine-tuning Llama2 with composable FSDP & PEFT methods to cover single/multi-node GPUs. Supports default & custom datasets for applications such as summarization & question answering. Supporting a number of candid inference solutions such as HF TGI, VLLM for local or cloud deployment.Demo apps to showcase Llama2 for WhatsApp & Messenger
Implementation of different applied machine learning algorithms from scratch with theory and details of implementation.
My take on deployment of ML model using FastAPI, Docker and Kubernetes.
Using traditional image processing techniques to construct 3D point cloud of objects. Incremental Structure from Motion (SfM) is used, a popular SfM algorithm for 3D reconstruction for reconstruction. The method is then evaluated using certain 3D reconstruction datasets.
Algorithm to texture 3D reconstructions from multi-view stereo images
Official Documentation: