Nitesh Prasad's repositories
biceps-curl-counter
Biceps Curl Counter python app based on Mediapipe's Human Pose Estimation architecture.
positionkey
Multi-platform python package for mapping key strokes to mouse clicks.
amazon-review-sentiment-analyzer
Amazon Fine Food Reviews
car-resale-price-predictor
Predict Used Car Price
credit-card-lead-predictor
Hackathon conducted by AnalyticsVidhya | | Implemented using Deep Neural Network to achieve validation AUC > 0.8710 | | Used Tensorflow/Keras library | | Used Keras Tuner for hyperpaprameter tuning
english-digit-recognizer
Convolutional Neural Network implentation from scratch using Keras | | Achieved validation accuracy of 99.4% within 10K trainable parameters with the help of Learning Rate Scheduler | | Data Augmentation using Keras ImageDataGenerator
face-mask-detector
Keras/Tensorflow implementation to detect faces wearing masks | | Data from Kaggle | | Used pretrained MobileNetV2 with imagenet weights(Transfer Learning) | | Data Augmentation using Keras ImageDataGenerator class | | Result obtained was 89.42% as validation accuracy | | Used OpenCV for inference on webcam feed
face-recognition
OpenCV, Keras
fake-news-classifier
LSTM implementation
intruder-mobile-notification-system
On Progress | | Notifications will be sent on Telegram with intruder images | | Stay tuned for more
investi
To automate Equity Trading
ml_projects
Projects
nsepy
Python Library to get publicly available data on NSE website ie. stock quotes, historical data, live indices
object-detection-yolo
Implemented YOLOv3 Object Detection using Keras/Tensorflow | | Used pretrained weights on MSCOCO dataset with 80 classes | | TBD: OpenCV implementation for demonstration on video streams
physiotherapy-pose-monitoring-system
Help patients in rectifying incorrect physiotherapy methods
pynput
Sends virtual input commands
stackoverflow-tag-predictor
Implemented One vs Rest Logistic Regression classifier | | 'Body' was removed and only 'title' was kept in data cleaning step | | Result obtained was 0.59 as micro-F1 score | | Tested on SVM but was costly in terms of time | |