High-Precision Bounding Box Annotation for YOLOv8 — Soccer Ball Detection (~2500 Frames)

I need a highly detail-oriented Image Annotator to assist with the critical task of preparing a dataset for YOLOv8 object detection training. The primary responsibility is to meticulously draw bounding boxes around a single target object (the soccer ball) within a set of 2500 sequential video frames. Due to the camera height and lighting conditions, the target object is often very small. This task requires extreme precision, as even slight deviations will negatively impact our model's performance. Project Specifications: - Volume: 2500 frames (JPEG format, already extracted). - Classes: Exactly 1 class (ball). - Output Format: YOLOv8 PyTorch format (a .txt file for every .jpg with normalized coordinates). - Tools: Preference for CVAT or Roboflow. Key Responsibilities & Annotation Rules: - Strict Bounding Boxes: Draw tight bounding boxes around the ball in each frame with no unnecessary padding. - Motion Blur Handling: When the ball moves fast and creates a blurred streak, the bounding box must enclose the entire streak accurately. - Partial Occlusions (Strict): If the ball is partially covered by a player, draw the box ONLY around the visible pixels. Do not guess the shape or outline the hidden parts of the ball. - Complete Invisibility: If the ball is entirely behind a player, out of frame, or completely burned out by harsh floodlight glares (0 pixels visible), do NOT label it. The model must only learn from real pixels, not estimations. - Quality Control: You will be expected to perform self-checks and strictly adhere to these edge-case guidelines to maintain high standards of quality. Qualifications: - Proven experience with YOLO-format annotation and bounding box precision. - Familiarity with tools like CVAT or Roboflow. - Good time management skills and the ability to maintain consistency across thousands of sequential frames. - Willingness to complete a small paid test milestone (~200 difficult frames) before the full contract is awarded. How to Apply: Please do not send a generic cover letter. To apply, simply include: - Your total fixed price for the entire dataset of 2500 images. - Your estimated turnaround time. - A confirmation that you agree to start with a short paid test to verify precision. We are looking for individuals, not agencies. Please only apply if you have hands-on experience with CVAT or Roboflow