Being a Pedestrian in an Indian City…

…is putting your life at a huge risk.

After hurting my foot multiple times due to gaping holes in sidewalks, and looking at the condition of roads in general, I can claim that urbanization in India keeps pedestrians as its last priority. This claim, though does not have enough data to back it, has clear signs of it being true.

One signal is the reports like the following:

  1. “Study: Why Pedestrians In India Need To Be Very, Very Cautious”
  2. Killer Indian roads claim lives of 56 pedestrians daily

There are hardly any decent sidewalks in India (barring some elite areas in New Delhi). In an Indian city that saw the fastest growth in late 1990s and early 2000s, where there is a decent sidewalk, there is something else that makes it unusable.

The following picture shows a footpath that stretches about half a KM, and in that stretch 4-5 such holes are left gaping for many days.

Since I prefer to walk and not drive, I have been encountering instances, where the unused spools of telecommunication or electrical cables are left on the pavements. In the picture below, the spool of electric cable is left laying there for over a month (or maybe even more) now. You can see the poor pedestrians walking on the newly constructed road (kudos for the new roads!).

I have observed several instances where the motorbike riders ride their bikes on the pedestrian pavements to beat the traffic.

Advertisements

You know what and where you are doing if you are using these flags

cmake -D CMAKE_C_COMPILER=/usr/bin/gcc-7 -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr -D OPENCV_EXTRA_MODULES_PATH=../opencv_contrib-4.0.1/modules -D BUILD_opencv_python2=OFF -D BUILD_opencv_python3=ON -D INSTALL_PYTHON_EXAMPLES=ON -D INSTALL_C_EXAMPLES=ON -D OPENCV_ENABLE_NONFREE=ON -D BUILD_EXAMPLES=ON -D BUILD_JAVA=ON -D WITH_TBB=ON -D WITH_EIGEN=ON -D WITH_VTK=ON -D WITH_OPENCL=ON -D WITH_OPENGL=ON -D WITH_CUDA=ON -D WITH_1394=ON -D WITH_CUFFT=ON -D WITH_CUBLAS=ON -D WITH_FFMPEG=ON -D WITH_GSTREAMER=ON -D WITH_INF_ENGINE=ON -D WITH_GDCM=ON -D WITH_QT=ON -D WITH_V4L=ON -D WITH_OPENCL_SVM=ON -D BUILD_opencv_cudacodec=OFF -D CMAKE_EXE_LINKER_FLAGS=-lcblas

OpenCV 4.0.1 deb Package for Raspberry Pi 3B+

It was a hellish experience building OpenCV4 (with NEON and VFPV3 support) without the cross-compile toolchain on a more capable hardware, with all required functionality (free, non-free, contrib etc.). But, I eventually did it… On a Raspberry Pi!

So that anyone else does not have to go through the ordeal over and over again, I am linking the DEB package I built for the Raspbian Stretch (Nov. 2018) here. Please check the SHA512 checksum against this. Enjoy!

SHA512 Checksum of opencv_4.0.1-1_armhf.deb:

7d75af63a779e014c22617a907adc32788587fde2d9029f4697b18c5556120d18cc5616c550a77157e80151cae4c7b19cd5f0cd876725063d93607506c0e1a06

I was short on time, so the package does not list what it provides (but believe me, it provides everything that can be done with OpenCV on 3B+ without TBB), and does not list what it depends on, so assume that it needs a lot of packages (none of them custom built, all installed binaries from official repos).

If it breaks anything for you, you are responsible!

Since I have been experimenting a lot with the so called #EdgeAI (is the term coined by Intel?) of late, I shall document my hacks to get OpenVINO + Intel NCS2 running really heavy models on Raspberry Pi satisfactorily. Clearly, other than the kick my kid gets out of it, my satisfaction doing this on the side comes from the fact that I could run a lot of ML theory in code on Pi.

Understanding CS Theory Research

I shall start posting what I could make out of the recent research articles from theory conferences and journals of my interest next weekend onward. The intention is to keep a public record of my understanding with a possible feedback from anyone interested in updating any shortcomings of my attempt. Each week, I hope to have covered 3 papers that I feel will prove to be seminal, at least in terms of the core idea.

My current interest areas are Cryptography, Computational Complexity Theory, Information and Coding Theory, Number Theory & Computational Algebra, Computational Geometry (Topology) and (Machine Learning only if something as interesting as Capsule Networks or GANs appears). I hope to learn a lot during this journey (for as long as I can carry on), so I am very excited.

If I find readers who engage in online discussions (and hopefully avoid phone and face-to-face conversation), I will start taking votes on what to read next week.

Trees

Most trees you’ll see in Computer Science literature are rooted on top and spread downwards. Why do CS folks call them tree then? Well, common wisdom is that they never went out of the room, so they never saw a real tree.