Table Tennis Stroke Detection and Classification

Setting up the project

1. Download the data (33gb) from the email.(Private)
2. Clone https://github.com/ultralytics/yolov5 into the present working directory
3. Add a new folder "imagedata"
   1. Add two more folders "images" and "labels" (imagedata/images, imagedata/labels).
   2. Add "train" folder under both images and labels (imagedata/images/train, imagedata/labels/train)
   3. Put all the labels from CVAT into the imagedata/labels/train folder
4. Run main.ipynb.
5. Create image.yaml in the yolov5 directory.
   1. Add the below contents to the file

path: imagedata/
train: images/train
val: images/train
nc: 1
names: ["stroke"]

6. Then run:
   1. cd yolov5
   2. For training: python train.py --img 1024 --batch 1 --epochs 1 --data image.yaml
   3. For detection (use any video): python detect.py --source ../data/videos/<video name>.mp4 --weights best.pt
7. The output video can be viewed in the folder ./yolov5/runs/detect/exp*