Background I am facing a lot of web and app that provide poor UX, especially they are difficult to be used for those who are not familiar with using web/app (e.g. elderly like my mum who don’t even get familiar with usual web/app navigation or what icons they represent) There have…

Background While haven’t logging in to the VM that host my local Kind Kubernetes cluster, I got an error connecting my cluster through kubectl, which is pretty weird for me. The error I got is as follow: The Discovery The first fews thing in my mind would be checking if the kubernetes node…

Intention The intention of this article is NOT to walk through in steps the steps manner, but focus on the gotchas I faced and an unexpected performance result. Setup I have a conda environment that have most required package installed, and the most important one is upgrade to PyTorch 2.0.1 (I tried…

Background I have had 2 different object detection projects on hand lately, and the manual annotation of data from scratch have been a frustrating task One platform I used is the website MakeSense, that’s a very good project that the image annotation is run on browser directly. The problem of such…

Background Given the previous experience on using YOLOv8 on Android through converting it from ONNX exported => Tensroflow => Tensorflow Lite (plus the realization on ditching tflite support object detection API and use plaint interpretor), there are following items that I am wondering if directly using PyTorch mobile would be a…

Background In my previous article, I have been working on converting the Ultralytics YOLOv8 nano model into ONNX model with adding of non max suppression (NMS)operation, by “stitching” a ONNX NMS graph node in the original YOLOv8 nano converted model, plus adding another PyTorch model that particularly transform the “selected indices”…

Background Following my previous post on exploring YOLOv8, I have been stuck at using the YOLOv8 model other than PyTorch, because the direct export model give result of dimension like [batch size, 5, 8400], which does encapsulated the result of overlapped bounding boxes and confidence score. But using this result in…

Background I have been exploring to train an computer vision object detection model and put into an Android device, the last time I have been exploring YOLOv7: Journey putting YOLO v7 model into TensorFlow Lite (Object Detection API) model running on Android Forewordsmedium.com Given YOLOv8 is out, I would like to see if there are any benefit to use YOLOv8 instead of YOLOv7. The flow and criterions

Background Following my previous post on inspecting backtrace message to see why the ESP32 crashed and reboot: Inspebacktrace / stack trace after ESP32 using espressif xtensa toolchain Backgroundstephencowchau.medium.com It’s very ineffective to trace 1 memory address at a time, so I wrote a powershell script for doing so, script as follow: To run the script, get to the location where the firmware.elf exist, which for PlatformIO, it’s your default Project folder > Project Name > .pio > build > (your env name, likely the board name, e.g. “nodemcu-32s”)