Ashish Papanai's repositories
stockDL
A financial deep learning library for stocks price prediction and comparison with traditional investment strategies. The Library is based on LSTM-Neural Networks and Conv1D + LSTM Neural Networks. Investments are subject to market risks, The AUTHOR HOLDS NO RESPONSIBILITY for any financial loss.
chessJS
This engine has defeated Stockfish Level-6 which has 2300+ ELO rating with 1sec as thinking time. It is an implementation of Chess Engine in JavaScript by using Min Max Tree with Alpha-Beta pruning (GOFAI (Good Old-Fashioned Artificial Intelligence)). The front-end of the website is basic HTML and CSS. ChessJS Version 2 : v2.chessjs.tech
al-folio
A beautiful, simple, clean, and responsive Jekyll theme for academics
stalemate
A UCI Chess Engine written in Python (Under Construction)
CabShortestPath
Reinforcement Learning solution to find the shortest path between the cab driver, user and the destination to minimise the faults and maximise the profit of the driver, by reducing travel time.
CDistNet-OpenVINO
Official Pytorch implementations of CDistNet
faceDetect
Python script to detect and extract faces from images to create specific datasets. It is based on the MTCNN Library which is an implementation of the ZHANG2016 research paper.
IMGprove
A deep-learning solution to improve the quality of low-resolution images. Improves the resolution of a 100x100 image to 400x400. This Library helps in optimizing data storage in cloud based servers without compromising much with the quality of the image. An Implementation of SRGAN (arXiv:1609.04802)
Inverted_Or_Not
Binary image classification model with 99.50 training accuracy, 98.20 Validation Accuracy and 98.34 testing accuracy on inverted and non-inverted human faces.
parseq-ov
Scene Text Recognition with Permuted Autoregressive Sequence Models (ECCV 2022)
pyprobml
Python code for "Machine learning: a probabilistic perspective" (2nd edition)
SpotX
Blocking ads and updates for the desktop version of Spotify, disabling podcasts and something else.
stockDL_MISP22
The stockDL algorithm assimilates two traditional stock trading techniques, Buy and Hold strategy and Moving Average ribbon trading strategy, with two Deep Learning Models created using the state-of-art Long Short-Term Memory networks. The first model is a pure LSTM network, whereas the second network is a Mixture of Convolution Neural networks and LSTMs. stockDL uses the data of the past five years from the date of generating the predictions, making the model immune from any sudden fluctuations in the historical data. When evaluated on the four stock symbols (AAPL, GOOGL, HDFCBANK.NSE, RELIANCE.NSE), the model has attained state-of-art for deep learning backed algorithmic trading in a controlled computational environment. The novel solution introduced in this study is faster and more accurate than any existing deep-learning solutions available. It is immune from any sudden dramatic decline among significant sections of the stock market. This work contributes to the stock analysis and research community of technical and financial domains.
VisionLAN-OpenVino
A PyTorch implementation of "From Two to One: A New Scene Text Recognizer with Visual Language Modeling Network" (ICCV2021)
ViTOL
ViTOL