Perception CU-ICAR AUE 8930

Final presentation and demo

1Record a demo video before the calss

2Mon(Apr.29) and Wed(May 1)

30min for each group.

3 submit code and presentation and report

Review of Previous class

  Humand vs deep learning

Pedestrian Detection:behavior

...

Outline

  Semantic segmentation/annotation, GAN image synthesis, train CNN with synthetic data, Imitation learing.

  1 Milestones

1943-Neural networks;

1957-1998;

2006-2018: dbn, IMAGE NET, ALEXNET, DROPOUT, GANs, deepface, bert.

Tools:

CUDA,caffe 2014(C++)  facebook, Tensorflow, Pytorch

Obeject detection

R-CNN--Overfeat-multibox,fast R-CNN-yolo-SSD-FEATURE PYRAMID NET(fpn)=reTINAnET-mASK R-CNN-FefineDet-M2Det

M2Det:https://arxiv.org/pdf/1811.04533.pdf

 

DeepLab family 

fully convolutional networks(fcns) feplace fully-connected layers with convolutional layers.

conditional random fields(CRFs) to caputre local and long-range dependencies within an image to refine the prediction map.

Astrous rate:percepiton of view

Semantic segmentation

 Annotation with Polygon-RNN++: interactive annotation tool, annotate your datasets much faster

semi supvised,

GAN:generative adversarial network ,from supervised--semi-unsupervised training.

   see:https://skymind.ai/wiki/generative-adversarial-network-gan

NVIDIA 2018. synthetic views

GAN image synthesis: synthesis fro label to image 

   https://tcwang0509.github.io/vid2vid/

  https://tcwang0509.github.io/vid2vid/images/teaser.gif

 Train CNN with Synthetic Data

    domain randomization;

    flying distractors:

https://arxiv.org/pdf/1804.06516.pdf

DAWNBench benchmark suite: increase the learning rate and decrease moment.

https://images.nvidia.com/content/tegra/automotive/images/2016/solutions/pdf/end-to-end-dl-using-px.pdf

 

Reinforcement Learning (RL)  https://en.wikipedia.org/wiki/Reinforcement_learning

compared with different machine learning:https://towardsdatascience.com/types-of-machine-learning-algorithms-you-should-know-953a08248861

cart pole problem:https://towardsdatascience.com/cartpole-introduction-to-reinforcement-learning-ed0eb5b58288

https://groups.csail.mit.edu/locomotion/

 

DL: general purpose framework for representation learning

   gieven an objective, learn representation,directly from raw input

RL:for decision making

RL:defines the objective

DL:gives the mechanism

RL+dl =GENERAL INTELLIGENCE

RL fundamentals:

  problems: policy leraning, reawrad value function learning; environment model learing(know current sate and action, predict next sate and reward)

charactersitics:close system with limited staes and actions;no direct rules/instrucations to action, you don't apply any,,

Problem formulation:

   experiment(observations, actions, rewards)

   sate is decied by experience

  in a fully observed environment: markov assumption  st=f(ot)

https://en.wikipedia.org/wiki/Markov_decision_process

value based RL

https://www.researchgate.net/post/what_is_the_difference_between_value_iteration_and_policy_iteration_methods_in_reinforcement_learning

https://web.stanford.edu/~anayebi/projects/CS_239_Final_Project_Writeup.pdf

https://selfdrivingcars.mit.edu/deeptraffic/

 

https://www.tesla.com/autopilot