comma Coding

 

Introduction: comma.ai, comma two (CTwo), *openpilot (OPGit, OP1, OP2, OP3, OP4)

Machine Learning (ML): AIabc, CNN, RNN

 

Homework YP
HW1:
Install Ubuntu 16.04 (must be 16.04): Guide1T, Guide2V. Use Ubuntu: Guide3, Guide4.   
HW2:
Install comma: Step1.txt ( Step1.m4v), Step2.txt (Step2.m4v), UI.
HW3:
Run view_steering_model.py: (A) Download research-master0.tar.gz and Data 10 to “Home” directory. (B) Go to “research-master0”. (C) Do Step3.txt (Step3.m4v). (D) Read view_steering_model.py, PTransform(), PT2, Keras1.
HW4: Run train_steering_model.py: (A) Move server.py to server1.py. (B) Download server.py. (C) Do Step4.txt (Step4.m4v). (D) Read train_steering_model.py, Keras2 (Seqtl), Keras3 (FAPI). (E) Read server.py.
HW5: Run and do JinnA (OPNet, Leon). Read LeonB, Shen; Keras4 (KM1, KM2, KM3); ENets (2019).
HW6: Run and do TomT (Yolact). Read Tsang, code2, code3 (DLab).
--- Do not do HW7 in the same Ubuntu. Step 5 failed Step 1.
HW7: Install and run
SNPE (Step5.txt). Read DLC1, DLC2, DLC3.

 

Projects
Project 1: Steering and Lane Detection: (A) YP: 4 Nets: commaNet + NvidiaNet + AlexNet + VGG. (R) 3 Data: comma, Nvidia (2016), Berkeley (2017), 2018, 2017, 2016, 2018, 2018, 2014,   
Project 2: Deployment: (A) JL: OPforkV + forkA (ANet) + forkK + forkS (h52py) (B) YP: forkA + Shen + OPNet + ANet + NN.py + .pb + .dlc (2018, 2018) (R) TeslaNetV, Git1, Git2, Atom1, Atom2, 2020 (RL),  
Project 3: Control: (A) YP, DH, JL: OPT (OPT1) + Shen + OPNet + NN.dlc (B) DL, JL: OP2 + Shen + ANet + Tune + Panda (PD1) + CAN (CAN1, CAN2, isotp, PyCAN , 2020) (C) JL: S&G (OP, forkS) + CP (CP1, CP2, CP3, CP4) + DBC (R) PID1, PID2, Kalman,    
Project 4: Instance Segmentation: (A) TT, BH: TomT + NN.py + .pb + .dlc + Yolact + 2019 (B) TT, WY, CY: TomT + JinnA + Shen + Yolact + .keras + .dlc (C) MOTS (D) comma10k (R) Coco, Yolo, Feature, Canny, 2019, 2016, 2017, 2016, 2014, 2012,
Project 5: Lane Change: comma, 2019, 2017, 2017, 2017,  
Project 6: Localization: comma2k19 (2018), Laika, 2019, 2018, 2017, 2014,   

 

1.         comma.ai

l  research: Code and data for 2016 openpilot

l  comma2k19: A driving dataset for the development and validation of fused pose estimators and mapping algorithms

n  raw_readers.ipynb

n  Data Type: Pose (Time, Position, Speed: ECEFEuler and Tait-Bryan AnglesLocal Tangent PlaneQuaternion), Steering, Path, Logs, Frame Reader.

n  Data Source: CameraRadarGPSIMUGNSSCAN

l   laika: Python GNSS processing library 

n  GNSS ProcessingTrilaterationLeast SquaresKalman Filter,

l  comma10k: 10k crowdsourced images for training SegNets

2.        comma two: comma Q&A, medium

l  (A) com1, com2, com3 (B) com4 (port1X, port1K): WB1 (run), WB2 (download), WB3 (install, WB4), SSH (key), port2, (C) fork1B, fork2B, fork3G (D) add1

l  Toyota Car: TSS18, TSS16, Sensors

3.        openpilot: Open source driving agent, OP5, OP6, OP Q&A, FAQ

l  openpilot-tools: openpilot development tools

l  opendbc: Democratize access to car decoder rings

l  panda: The nicest universal car interface ever

n  Introduction to embedded systems

n  STM32CubeIDE (15 m); STM32 Slides-Part1, Part 2; STM32 Book1 (25 p); STM32 Book2 (125 p); STM32 Book3 (416 p)

u  CMSIS Step 1 (8 m); STM32 Arduino IDE (4 m); STM32 MDK-ARM IDE (8 m)

n  CNN on STM32 (6 m), Github, Guide; STM32a; STM32b (CMSIS-NN); STM32c (Deploying); Paper

4.        OP: Makefile, package, wheel, docker, __init__,

5.        Python: Code1, Tutor, argparse, collections, datetime, gc, h5py, json, multiprocessing, numpy, os, pygame, scikit-image, signal, sys, threading, time, uuid, zmq,

6.        ML: 10CNN, ResNet,