[Real-time detection without GPU](https://youtu.be/l3xqD581TQc)
cuda_11.1
python version- 3.7.13
Data preparation
Using the reference materials, you can find a link to the dataset. You can preprocess dataset in roboflow. As you know YOLOv5 еakes image size 416x416, so I resized it to this size. To use the dataset you need to click as in the following photos:

Train the model and detect cards
To see how to define model configuration and architecture, train and detect model you need to clone my repo and open file "Train_Yolov5.ipynb". In folder weights I saved my post-train weights. You can use them or save your.Sample result(trainbatch):

Dependence of metrics on the number of epochs:

Real-time webcam detection
To see how to define model configuration and architecture, train and detect model you need to clone my repo and open file Real_time_webcam_Yolov5.ipynb. This notebook is using file webcamdetect.py where you may find necessary functions.References
-
https://public.roboflow.com/object-detection/uno-cards - Uno cards dataset;
-
https://github.com/ultralytics/yolov5 - Original repo of YOLOv5;
-
https://models.roboflow.com/ - Model zoo from roboflow;
-
https://www.youtube.com/watch?v=nDPWywWRIRo&t=3256s&ab_channel=StanfordUniversitySchoolofEngineering - Basic Object Detection knowledge;
-
https://www.youtube.com/watch?v=MdF6x6ZmLAY&t=1508s - Yolov5 tutorial;
-
https://www.youtube.com/watch?v=NU9Xr_NYslo&t=607s - Yolov5 tutorial;
-
https://www.youtube.com/watch?v=yfDjsuxIKA4&t=2718s - Training other models using Tensorflow Object Detection;
-
https://www.youtube.com/watch?v=pnntrewH0xg&t=151s - Example of web-app for testing your model;
-
https://www.youtube.com/watch?v=TB-fdISzpHQ&t=3717s - Another Basic Object Detection knowledge;
-
https://towardsdatascience.com/yolo-v4-or-yolo-v5-or-pp-yolo-dad8e40f7109 - Difference between the last YOLO-type models;
-
https://techzizou.com/category/object-detection/ - Web app on tf2;
-
https://github.com/tensorflow/models/blob/master/research/object_detection/g3doc/tf2_detection_zoo.md - model zoo(tf2);