Skip to content

Add automatic gameplay data collection and Windows build fixes#54

Merged
SunOner merged 2 commits into
SunOner:mainfrom
asenyeroao-ct:data-collector-debug-updates
May 22, 2026
Merged

Add automatic gameplay data collection and Windows build fixes#54
SunOner merged 2 commits into
SunOner:mainfrom
asenyeroao-ct:data-collector-debug-updates

Conversation

@asenyeroao-ct

@asenyeroao-ct asenyeroao-ct commented Mar 21, 2026

Copy link
Copy Markdown
Contributor

Summary

This PR adds automatic gameplay data collection for YOLO training and includes the Windows project fixes needed to build and run it in the current setup.

The app can now save gameplay frames during normal use and optionally generate YOLO-format .txt label files from the active detector output. It also exposes the full workflow in the Debug overlay so collection can be enabled, tuned, and monitored at runtime.

What it does

  • adds a data_collector flow for saving gameplay frames as JPEG samples
  • groups collected samples by model name under the resolved output folder
  • optionally writes YOLO label files using the current detection boxes, classes, and confidences
  • supports filtering labels by confidence threshold, max boxes, and selected class IDs
  • adds Debug UI controls for enabling collection, frame interval, JPEG quality, output folder, and auto-label options
  • shows runtime counters, resolved output path, and last status directly in the overlay

Build and project updates

  • updates the Visual Studio project so CUDA build customizations can resolve more reliably on Windows
  • avoids the old serial.lib toolset mismatch by compiling the serial sources directly in the main project
  • keeps the new collector files included in the Visual Studio project and filters

Runtime behavior

The collector can be configured to:

  • collect only while aimbot is active
  • collect only when targets exist
  • save only every N frames
  • save images only or images plus YOLO labels

A cooldown is also applied to avoid bursts of nearly identical samples.

Verification

  • verified locally by running the application and opening Overlay -> Debug -> Data Collection
  • confirmed the new Debug UI is visible and configurable at runtime
  • confirmed the resolved output path is shown in the overlay
  • confirmed the application runs with the integrated project changes in the current environment

@asenyeroao-ct asenyeroao-ct changed the title Add data collector and detector debug updates Add automatic gameplay data collection for YOLO training Mar 21, 2026
@asenyeroao-ct asenyeroao-ct changed the title Add automatic gameplay data collection for YOLO training Add automatic gameplay data collection and Windows build fixes Mar 21, 2026
@phenomen4uCZ

Copy link
Copy Markdown

Would it be possible to add support for the rp2040 or rp2350 device? Thank you for your consideration...

@SunOner SunOner merged commit c40dc01 into SunOner:main May 22, 2026
SunOner added a commit that referenced this pull request May 22, 2026
Add the new data_collector.cpp source file to the CMake source list so DML CMake builds link successfully after PR #54.

Update the Visual Studio DML project settings to use the prebuilt OpenCV paths used by the setup script, and remove stale GLFW link dependencies that are no longer used by the project.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants