Back to Projects

YOLO Fishing Bot

Primary category
AI / Computer Vision
Supporting categories
  • Machine Learning
  • Automation
  • Python

Overview

I manually labeled 3,350 game-interface images and trained a custom object detector on my PC. I then wrote a Python Fishing Bot that uses YOLO detections to trigger fishing inputs in the game.

My Role & Contribution

I handled dataset labeling, model training, and the Python automation application. The preserved runtime source and checkpoint support the application and model lineage, while the frozen Owner record supplies the training and authorship boundary.

Technical Approach / Technologies

The detector was fine-tuned from a pretrained YOLOv8m starting checkpoint rather than trained from scratch. At runtime, the Python application captures the game window, runs a 42-class detector, interprets selected detections, and issues corresponding F1/F inputs.

Technologies

  • Python
  • Ultralytics YOLOv8
  • PyTorch
  • OpenCV
  • Win32 screen capture
  • PyDirectInput
  • custom checkpoint

Engineering Decisions / Trade-offs

The application uses window capture and local inference so detection and input decisions occur on the same PC. A custom 42-class checkpoint supports multiple interface states.

Challenges / Lessons

Preserved frames show difficult interface conditions such as occlusion, dense overlays, and duplicate detections. They support qualitative discussion rather than a performance measurement.

Results / Status

The preserved checkpoint and recording show a custom detector producing multi-class predictions, and the source connects selected predictions to automated fishing inputs. Preserved summary values remain internal rather than serving as public performance claims because their evaluation context is incomplete.

Evidence

The case study is supported by the pinned Python repository, checkpoint metadata, a 13-minute-plus inference recording, extracted frames, and the frozen Owner confirmation for dataset and training authorship.

Training provenance and runtime flow for the YOLO Fishing Bot.

Explanatory diagram separating local fine-tuning from runtime automation. Alan manually labeled 3,350 images and fine-tuned a pretrained YOLOv8m checkpoint; the Python application connects detector output to fishing inputs. This is a diagram, not an application screenshot or performance measurement.

View full-size diagram