Refer to Launchpad PPA:
1 | dput ppa:longervision/googletest googletest_1.11.0_amd64.changes |
~/.dput.cf as:1 | [googletest] |
NEVER successfully built out QtWebEngine before. Today, I’d love to have a try. From this tutorial Building QtWebEngine, it looks possible.
Nothing but a collection of various tutorials.
Happened to come across a blog XGBoost vs LightGBM: How Are They Different. Let’s investigate a bit wider and deeper into the following 4 machine learning open source packages.
Cited from XGBoost vs LightGBM: How Are They Different
1 | Two of the most popular algorithms that are based on Gradient Boosted Machines are XGBoost and LightGBM. |
Unlike voting or stacking ensembles, which are multiexpert systems, cascading is a multistage one.originated in the observation by Leo Breiman, an American academician, also proposed Random Forest.In this blog, I’m going to use a very old Kaggle dataset Amazon.com - Employee Access Challenge.
Please make sure:
1 | ➜ ~ pip show xgboost |
Please make sure:
1 | ➜ ~ pip show lightgbm |
Pay Attention: In order to have XGBoost or LightGBM successfully built, you may have to:
1 | ➜ ~ pip show catboost |
1 | ➜ ~ pip show h2o |
1 | ➜ pip show mlflow |
Please make sure:
Directly refer to Rapid-Fire EDA process using Python for ML Implementation.
Clearly, PCA is a part of EDA.
Here is another good tutorial for EDA: A Starter Pack to Exploratory Data Analysis with Python, pandas, seaborn, and scikit-learn.
Let’s just refer to the FIRST tutorial and try out how to analyze and visualize the famous dataset MNIST.
Please refer to my copy-paste-modification jupyter notebook code at https://github.com/LongerVision/Kaggle.
Let’s then refer to the SECOND tutorial and try out how to analyze and visualize Fashion-MNIST.
Please refer to my copy-paste-modification jupyter notebook code at https://github.com/LongerVision/Kaggle.
Happy new year everyone. Such a long time I haven’t taken a look at activation functions. It might be a good time for me to review and check.
Let’s first take a look at activation functions of year 2022.
I would strongly recommend you to read the 2 articles mentioned above, particularly: Smish.
It’s been over a year and a half for JeVois to have this second generation of smart camera JeVois Pro. Although JeVois failed to pledge its campaine goal on Kickstarter, it’s still my great pleasure to give it a try. Again: JeVois Pro is made of China.
| Specifications | JeVois Pro | JeVois |
|---|---|---|
| CPU | Amlogic A311D with 4x A73 @ 2.2 GHz + 2x A53 @ 1.8 GHz | Allwinner A33 quad core ARM Cortex A7 processor @ 1.34GHz with VFPv4 and NEON |
| GPU | Quad-core MALI G52 MP4 @ 800 MHz | dual core Mali-400 GPU supporting OpenGL-ES 2.0 |
| NPU | 5-TOPS (trillion operations/s) integrated Neural Processing Unit. | None |
| RAM | 4 GB LPDDR4-3200 | 256MB DDR3 SDRAM |
| Camera | 2MP Sony IMX290 back-illuminated Starvis sensor, 1/2.8”, 12mm lens, 1920x1080 at up to 120fps, rolling shutter, wide dynamic range support. | 1.3MP |
| IMU | TDK InvenSense ICM-20948 with 3-axis accelerometer, 3-axis gyro, 3-axis compass, SPI bus @ 7 MHz, can be synchronized with camera sensor. | None |
| … | … | … |
JeVois Pro and JeVois respectively look like the following:
Now, let’s follow the Quick Start and take a test of JeVois Pro. After successfully flashing the TF card, let’s boot up JeVois Pro, For the first booting, it’s better we connect both HDMI and keyboard and mouse.
JeVois Pro has 3 booting choices, as follows.
For my convenience of setting up the network, I chose Ubuntu Graphical. And, now it’s the time for us to plug in our Wifi dongle/module before rebooting. Then, reboot.
Clearly, JeVois Pro‘s Ubuntu Graphical comes with GUI XFCE.
1 | ➜ ~ ssh jevois@192.168.1.78 |
According to my Email to JeVois Inc. and their reply, JeVois Pro hasn’t implemented the video streaming part yet. Clearly, two streams are needed to be transferred via Wifi:
Let’s leave this issue aside for now…
Merry Christmas everyone. It is another Christmas, Year 2022 is coming. I’m back to life. Last week, my JetBot arrived. It took me a couple of hours to make it work, and round 1 week to demonstrate: some of the JetBot Examples are NOT working properly. Anyway, my son Fionn is half-year old. Let’s celebrate it with JetBot, as well as the heavy Christmas snows.
| Christmas Snow | My Son | Snow |
|---|---|---|
 |
 |
 |
 |
 |
 |
 |
 |
 |
I believe we can flash JetPack first and have JetBot configured afterwards. Let’s leave this topic for the next blog The JeBot - 2. In this blog, we’re going to follow the official JetBot Software Setup.
Just refer to this youtube video:
Since what I purchased is a Waveshare JetBot 2GB AI Kit. therefore, my assembled JetBot looks like following:
As you can see, there is NO antenna. Instead, a Wifi USB dongle is provided. Actully, there are a couple things to be emphasized for my Waveshare JetBot 2GB AI Kit:
We just download the JetBot official SD card image from Software Setup (SD Card Image), and flash it to my 64G TF Card.
1 | ➜ jetbot sudo dd bs=1M if=jetbot-043_nano-2gb-jp45.img of=/dev/mmcblk0 conv=fsync |
After that, as usual, we upgrade the entire JetBot system by sudo apt update and sudo apt upgrade. After all these, our JetBot looks like:
1 | ➜ ~ python --version |
After twickling around how to run pip under /home/jetbot/.local/bin for user root, and have Jetson Stats installed using this command sudo -H pip install -U jetson-stats, we got the following results:
Anyway, let’s take a look at my installed results:
Do NOT forget to export a couple of environmental variables in both .bashrc and /etc/environment as:
1 | export OPENBLAS_CORETYPE="AARCH64" |
It looks that by directly installing pytorch via pip install torch -U --user, a second libgomp will be installed. Both libgomps brought me the error of cannot allocate memory in static TLS block, please refer to:
The solution to the above errors is:
Pretty much similar to:
JetBot Basic Motion Demo is really a very cool demo for you to start with JetBot. It wouldn’t be hard for you to run a Jupyter Notebook Server remotely. Anyway, my testing results are given as follows: a headache video, as well as .pdf summary.
1 | ➜ sudo dd bs=1M if=/dev/mmcblk0 of=jetbot.img conv=fsync |
Finally, I attached my built JetBot OS at https://www.longervision.us/iso/jetbot-nano2g-ubuntu18.04.6-cuda10.02-cudnn8.0.0-opencv4.5.4.img. You are welcome to take a try.