REU - End to End Self Driving Car

Slides:



Advertisements
Similar presentations
Presentation by: Serena, Ann & Nicole
Advertisements

Patch to the Future: Unsupervised Visual Prediction
Standard Grade Automated Systems and Industrial Applications Automated Systems and Industrial Applications Standard Grade.
Computer Architecture and Organization Introduction.
Outline What Neural Networks are and why they are desirable Historical background Applications Strengths neural networks and advantages Status N.N and.
Chapter 1 Introduction. Architecture & Organization 1 Architecture is those attributes visible to the programmer —Instruction set, number of bits used.
Slide 1 Welcome to SimELMO; Newspaper (Print Front and Back) Edition! In this simulation you will learn to display newsprint pages that have text on both.
Feedforward semantic segmentation with zoom-out features
Sparse Coding: A Deep Learning using Unlabeled Data for High - Level Representation Dr.G.M.Nasira R. Vidya R. P. Jaia Priyankka.
FINGERTEC FACE ID FACE RECOGNITION Technology Overview.
S.Bengio, O.Vinyals, N.Jaitly, N.Shazeer
When deep learning meets object detection: Introduction to two technologies: SSD and YOLO Wenchi Ma.
Unsupervised Learning of Video Representations using LSTMs
Lesson Objectives Aims You should be able to:
Week 3 (June 6 – June10 , 2016) Summary :
Recurrent Neural Networks for Natural Language Processing
Deep Predictive Model for Autonomous Driving
Textual Video Prediction Week 2
Summary of Week 1 (May 23 – May 27, 2016)
Unit 3 – Driver Physical Fitness
A Hierarchical Model of Reviews for Aspect-based Sentiment Analysis
Part-Based Room Categorization for Household Service Robots
Synthesis of X-ray Projections via Deep Learning
CNN Demo LIU Pengpeng.
AV Autonomous Vehicles.
Autonomous Cyber-Physical Systems: Autonomous Systems Software Stack
Unsupervised Learning and Autoencoders
ASAP and Deep ASAP: End-to-End Audio Sentiment Analysis Pipelines
Deep Learning and Newtonian Physics
Textual Video Prediction
Image Question Answering
Monte Carlo Simulation Managing uncertainty in complex environments.
Introduction to Deep Learning for neuronal data analyses
Simon Hecker, Dengxin Dai, and Luc Van Gool
A Comparative Study of Convolutional Neural Network Models with Rosenblatt’s Brain Model Abu Kamruzzaman, Atik Khatri , Milind Ikke, Damiano Mastrandrea,
Two-Stream Convolutional Networks for Action Recognition in Videos
A First Look at Music Composition using LSTM Recurrent Neural Networks
Recurrent Neural Networks
The Open World of Micro-Videos
The Story of Jacy Good.
Basics of Deep Learning No Math Required
Word embeddings based mapping
Word embeddings based mapping
Papers 15/08.
Object Classification through Deconvolutional Neural Networks
Deep Cross-media Knowledge Transfer
Recurrent Encoder-Decoder Networks for Time-Varying Dense Predictions
Deep Neural Networks: A Hands on Challenge Deep Neural Networks: A Hands on Challenge Deep Neural Networks: A Hands on Challenge Deep Neural Networks:
边缘检测年度进展概述 Ming-Ming Cheng Media Computing Lab, Nankai University
Natural Language Processing (NLP) Systems Joseph E. Gonzalez
Attention for translation
Department of Computer Science Ben-Gurion University of the Negev
Lecture 21: Machine Learning Overview AP Computer Science Principles
Automatic Handwriting Generation
Week 8 Farzain Majeed.
Using Clustering to Make Prediction Intervals For Neural Networks
The experiments based on Recurrent Neural Networks
Presented By: Harshul Gupta
Week 9 Farzain Majeed.
Week 3 Presentation Ngoc Ta Aidean Sharghi.
CRCV REU 2019 Kara Schatz.
Week 3 Volodymyr Bobyr.
Self-Supervised Cross-View Action Synthesis
Bidirectional LSTM-CRF Models for Sequence Tagging
REU 2019 Week 2 Volodymyr Bobyr.
Week 7 Presentation Ngoc Ta Aidean Sharghi
LHC beam mode classification
Lecture 9: Machine Learning Overview AP Computer Science Principles
The experiment based on hier-attention
Presentation transcript:

REU - End to End Self Driving Car Farzain + Kishan

BEWARE

Quick Recap Create an end-to-end system that is able to autonomously maneuver a car through a variety of scenarios. Only allowed to use CAMERAS. Extremely complex situations. IMPOSSIBLE. Deep learning can help us out.

Related NVIDIA has released several reports on their work. Works very well for simpler situations (staying in lane, not hitting cars in front) Doesn’t work so well in more complex situations that are harder for a network to recognize (traffic signal, pedestrians) This is partially because there isn’t enough training data and partially because the network may have a hard time differentiating between whats important and what isn’t.

Related Princeton paper recently proposed an FCN-LSTM Wanted to use ONLY dash cam videos, nothing else, to predict future movement. Almost turned the problem into a text prediction problem. Semantic segmentation label side training to “learn a meaningful feature representation”. Here they are implicitly feeding it segmentation information. Doesn’t go directly from action -> input, “future motion action feasibility distribution”

Other

Udacity Labeled datasets. Challenged programmers to create a deep network to predict steering angles. First place used a 3D CNN + LSTM + RNN Second place used Optical Flow + NVIDIA CNN Top 5 models are open source.

Model 2D CNN DECISION

Model SEG LOSS FCN LSTM DECISION

Model 3D CNN LSTM DECISION

Where do we go from here?

Model 2D CNN

Model 3D CNN

Model 3D CNN LSTM

Model 3D CNN LSTM DECISION

Issues: output of two stream CNN, memory, usage of another model 3D CNN RGB 3D CNN LSTM DECISION Segmentation Model Output 3D CNN Issues: output of two stream CNN, memory, usage of another model Pre-training?

Output as probability distribution Model DECISION Output as probability distribution

Next 7 Days Code up/train NVIDIA base model in Keras (half-done) Code up Udacity dataset reader script (half-done). Get results of NVIDIA model to establish a solid base. Create 2D CNN + LSTM and see results