Video Crack Detection Model

Job ID: 39934171

Budget: $30 – $250 USD

I have a single, lengthy video and need a lightweight proof-of-concept model that can flag visible cracks whenever they appear. The task is focused and time-sensitive—ASAP delivery is preferred.

What I will supply
• The full video file (or a link to it).
• A shortlist describing the exact “crack” patterns I consider relevant.

What I need from you
• A trained object-detection model (YOLOv5/YOLOv8, TensorFlow, or PyTorch—use whichever you are fastest with) that can spot cracks frame-by-frame.
• An inference script that takes the video as input and outputs timestamps and, ideally, bounding-box coordinates for each detected crack.
• Brief setup instructions so I can run the script locally.

Keep the scope lean: no heavy dataset curation—use transfer learning and any open-source crack datasets you already know. The end goal is a working demo that shows reliable detection on my footage, nothing more at this stage.

If you can turn this around quickly with clear results, let’s get started right away.