January 31, 2020
A 102 category dataset consisting of 102 flower categories, commonly occuring in the United Kingdom. Each class consists of 40 to 258 images. The images have large scale, pose and light variations.\ I have used fastai library and pytorch for this project.\ I got accuracy of 97.84%\ For detailed explaination of code step by step visit my Blog
This my engineering final year major project along with 3 other team mates. We have published paper in International Research Journal of Engineering and Technology (IRJET) Volume 7, Issue 3, March 2020 S.NO: 924 Our published paper\ We have successfully shown that CNNs are able to understand the entire learning process lane and road following without manual decomposition into road or lane marking detection, path planning, and control.\ A small amount of training data from less hours of driving was sufficient to train the virtual car to operate in diverse conditions, on highways, local and residential roads in sunny, cloudy, and rainy conditions. The CNN is able to extract the meaningful and useful road features from a very sparse training signal(only steering).
This is Survival Estimates that Vary with Time.\ Analyze surivival estimates for a dataset of lymphoma patients considering Censored Data and Kaplan-Meier Estimates\ For detailed explaination of code step by step visit my Blog
I used the Oxford-IIIT Pet Dataset by O. M. Parkhi et al., 2012 which features 12 cat breeds and 25 dogs breeds. My model will need to learn to differentiate between these 37 distinct categories.\ I got accuracy of 94% which is just 6% error with just few lines of code, when compared to state of art model in 2012 paper which had 56% accuracy.\ For detailed explaination of code step by step visit my Blog here
Evaluating Treatment Effect Models
Comparing predicted and empirical risk reductions, Computing C-statistic-for-benefit Interpreting ML models for Treatment Effect Estimation
Building a risk score model for retinopathy in diabetes patients using logistic regression.
This project is basically building SGD from scratch by learning all basics need to know.\ Even though we have high level api to do all the functions for us, still it is important to learn few important(if not all) concepts.\ SGD taking each step is shown using animation.
I have solved Kaggle problem House price regression using 2 methods.
- First method\ Traditional and most important way of solving, Apply machine learning techniques like preprocessing and all other feature engineering steps. And then applied random forest and also tried xgboost
- Second method apply neural network to solve problem. for this I tried fastai library
solved using fastai library