YOLOv4 | Port of YOLOv4 to C # TensorFlow | Machine Learning library

I am trying to determine if certain parts of an RGB image are colored or grayscale, using python, opencv and numpy libraries. To be more spesific, in an RGB image I determine face locations using neural networks and when that image contains printed photos I would like to find out if the face location in that image is grayscale or colored.

What I tried so far:

I have integrated OpenVINO and PyQt5 to do the inference job as shown in the image on Windows 11 with openvino_2021.4.689 version.

I reference this GitHub to finish YOLOv4 inference with NCS2.

The following is my inference engine code.

I am trying to install openvino_2021.4.689 version with Windows 7 old computer.

I need to use OpenCV with my project, so I have to use PowerShell to execute ffmpeg-download.psl in opencv folder of openvino_2021.4.689 like this.

If I install my own OpenCV by pip install opencv-python in Command Prompt rather than install by OpenVINO's ffmpeg-download.psl file, my project with cv2 library will not work successfully.

Specifically, my YOLOv4 frame will not show without any error message, what I use cv2 to draw images cannot work.

But if I click ffmpeg-download.psl with right mouse button and select executed by PowerShell, I will get an error message as shown in the image. (Executed with system administrator.)

My environment is Windows 11 with openvino_2021.4.752 version.

When I try to run object_detection_demo.py in the demos folder of inference engine, N/A result will be occurred using CPU, and the MYRIAD issue I will mention later will happened with my NCS2.

Before I am using RTX2070 SUPER to run Pytorch Yolov4 and now my PC is changed to use RTX3060, ASUS KO GeForce RTX™ 3060 OC.

I have deleted the existing cuda11.2 and install again with cuda11.4 and Nvidia Driver 470.57.02

I trained a model with yolov4. The inference is perfect and so are the metrics. I converted the model to tensorflow lite to be able to use it on android. I would like to view the accuracy, precision and recall values ​​of the converted model. How can I do?

What is the best way to run YOLOV4/YOLOV4-TINY on RPI 4 using Tensorflow-lite for object detection? I want to detect/count the no. of people in the room using this followed by detection of items like chair, banana e.t.c?

As far as I know these libraries have MIT license and can be used for educational/commercial purposes, is that correct?

Also, what works better on Rpi 4 with tensorflow lite, is it YOLOv4 or YOLOv4 Tiny or something else?

Thank you.

I am working with YOLOv4 for detection through IP Camera. I have a GUI for camera control. So I don't want the camera frame to show the detected objects. However, I want the detected objects and the percentage to be shown in the Command Prompt console. Is it possible to make that? If yes, please suggest the way. Thank You

I have done the model training for Yolov4 objection detection from AlexeyAB Darknet package on Colab. (from https://github.com/AlexeyAB/darknet) Two classes are defined named "scratch" and "blood". Then run the following commands for testing on a single image:

!./darknet detector test data/obj.data cfg/yolo-obj.cfg backup/yolo-object_last.weights images/meat/test/376.jpg -dont show

and the result is shown below. It's expected that there are only one scratch and blood shown with probabilities in the result. However, it shows so many scratch and blood predictions (the last few lines of this post)! The number of classes (=2), class names, and shoud have been set properly in obj.data and yolo-obj.cfg. Can anyone advice why and how to resolve it?