Manish Bhurtel

Housing Price Prediction and Home Equity Value Estimation

✓ Built Housing Price Prediction System using Random Forest on Zillow Data, and estimated the equity value using Amortization Table.
✓ Engineered the Zillow data to increase the R2-score from 75% to 88% on validation set
Tools and Technologies: Python, Pandas, Numpy, Scikit-learn, Matplotlib
Duration: Aug 2021 – Dec 2021

Code

Federated Learning (FL) Simulation on VANET Dataset

✓ Implemented FL on VeRiMi VANET dataset using TensorFlow and compared with the performance standard DL Method
✓ Achieved 3% greater accuracy using FL than the DL Method, trained using same backbone neural network
Tools and Technologies: Python, TensorFlow, Pandas, Numpy, Matplotlib
Duration: Oct 2020 – Nov 2020

Code

Background Motion Detection

✓ Leveraged Image Processing to detect "background motion", which passed the QA test with 95% accuracy (with 100 total test videos)
Tools and Technologies: Python, OpenCV, HSV Image Processing
Duration: May 2021 – Jul 2021

Code

2D Poster Injection using Computer Vision

✓ Automatically understanding the scene and injecting 2D images/posters at a perfect place onto the background of the video.
Tools and Technologies: Python, OpenCV, Dlib, NumPy, Matplotlib, Image Processing, Foreground Segmentation
Duration: Dec 2020 – Jul 2021

Demo Videos

Object Detection and Product Recommendation

✓ Detecting different items and if the user clicks on the detected items, it shows the list of related items that the user would prefer.
✓ Implemented YOLOv3 to detect the objects and used Cosine Similarity to recommend the matching items corresponding to the detected object.
Tools and Technologies: Python, Pandas, YOLOv3, NumPy, TensorFlow, Keras, OpenCV, Scikit-learn, Cosine Similarity
Duration: Sept 2020 – Nov 2020

Code

Hairstyle Recommendation System using Deep Learning and Computer Vision

✓ Prepared a Face Shape Dataset with faces belonging to 5 categories (oval, oblong, heart, round, square)
✓ Designed a custom Deep Learning Architecture to classify the face shape detected by Dlib Face Detector
✓ Created and overlayed hair templates onto the head position on the basis of detected face shape of the person
Tools and Technologies: Python, Dlib, CNN, OpenCV, Django, Photoshop
Duration: Aug 2019 – Oct 2019

Code

Lattice Constant Prediction of Cubic Perovskite Oxides

✓ Data engineering [preprocessing, anamoly removal, correlation analysis]
✓ Comparative analysis of different regression models to predict the values of lattice constant of Cubic Perovskite Oxides
Tools and Technologies: Python, Pandas, NumPy, TensorFlow, Sklearn, Matplotlib, Seaborn, Plotly, ANN, KNN, RBF kernel SVM, Decision Trees, Random Forest.
Duration: June 2020 – July 2020

Code

Deep Learning Based Seed Quality Tester

✓ Created a high-quality Maize Seed Image Dataset using fine quality seeds, damaged seeds, and foreign elements [link]
✓ Trained a customized 5-class Deep Learning Classification Model, and achieved 81% training and 80% validation accuracy
Tools and Technologies: Python, CNN, TensorFlow, Keras, Django
Duration: Aug 2018 - Dec 2018

Code

Face Validation

Automatically detecting the faces and accepting only valid face images in the person’s official profile. The valid photo contains a single faced person facing front without sunglasses or mask.
Tools and Technologies: Python, Dlib, TensorFlow, Keras
Duration: Jun 2019 - Jul 2019

Code

Abusive Content Detection in Text

Automatically detecting the textual posts which contains adult or abusive content. Those adult or abusive posts are automatically rejected.
Tools and Technologies: Python, NLTK, Naïve Bayes Classifier
Duration: Jun 2019 - Jul 2019

Code

HELLO WORLD