ujjwalsaxena Goto Github PK

automold--road-augmentation-library

This library augments road images to introduce various real world scenarios that pose challenges for training neural networks of Autonomous vehicles. Automold is created to train CNNs in specific weather and road conditions.

behavior-cloning-dataset-ujjwal

Behavior training data

behavior-cloning-t1p3

The project included designing a neural network and then training the car on the road. The CNN had to learn and clone the driving behavior.

carnd-advanced-lane-lines-detection-t1p4

This is the advanced lane detection project which includes advanced image processing to detect lanes irrespective of the road texture, brightness, contrast etc. I used Image warping and sliding window approach to find and plot the lanes. This makes it better with the lane curves.

extended-kalman-filter-t2p1

This includes Lidar and Radar data fusion. The radar measurement space being a non linear function requires linearization to apply Kalman Filter. This is done using Taylor series and Jacobian matrices in an Extended Kalman Filter approach.

functional-safety-project-t3p3

kindnapped-vehicle-t2p3

Localization

lane-line-detection-t1p1

In this project I used some Image processing techniques like blurring, Canny Edge detection, Hough transform etc. to detect the lane lines on the road and then plotted a lines over them to mark the lanes.

lstm-sentiment-analysis

Sentiment Analysis with LSTMs in Tensorflow

mpc-project-t2p5

Model predictive control project takes the waypoints as input and predicts a trajectory for driving vehicle accordingly.

path-planning-project-t3p1

Prediction, behaviour planning and trajectory

pid-control-project-t2p4

PID control

python-opencv-cuda

custom opencv_contrib module which exposes opencv cuda optical flow methods with python bindings

rabbit-dashboard

road-glare-removal-tests

semantic-segmentation-t3p2

Semantic segmentation uses transposed convolutions in the decoder part to segment the images in classes. This projects implements a FCN8 architecture to detect the roads in the image.

simple-eye-detection-tests-ml

traffic-sign-classifier-with-inception-layer-t1p2

This is the second project by Udacity's Autonomous vehicle development program. A CNN was designed and trained to detect the traffic signs. The system was also tested on German traffic signs to measure its performance.

udacity-carnd-capstone

udacity-image-classification

Udacity Data Science Nanodegree program, deep learning on Pytorch, image classification (flowers)

unscented-kalman-filter-t2p2

This is another approach for linearizing a function to apply Kalman Filter. Instead of linearizing only around mean, sample points are taken for each dimension and then a Gaussian is approximated around them.

vehicle-detection-t1p5

This project used machine learning to train a classifier to classify cars in video and then draw bounding boxes around them. I used sliding window technique, and heat map to bind the car images in rectangles.