Object Detection - Tennis Ball
Our data consists of 150 high-quality labeled images of tennis balls in different settings, that were manually taken from different cameras. The images vary in resolution, lighting conditions, distance from the ball, angles of the camera, the position of the ball with respect to the image, number of other objects present in the frame, etc. This makes the dataset robust for object detection needs.
Steps to reproduce
1. The data was acquired by manually taking images of tennis balls in varying environments in order to create a robust object detection model. 2. Once the images are collected, they are tested with the "test.py" Python script, which basically checks whether the image is processable or not and is not corrupted. If any image(s) are corrupted, then it deletes those images. 3. Once the images are verified, they are processed by another Python script, which renames all the images in the format “00000001” for the first image and “00000100” for the 100th image, and so on. 4. Then, the renamed images were manually labeled through an open-source software called LabelImg. Link: https://github.com/tzutalin/labelImg