My model correction is not improved even if i increased my dataset

🤝 Community Help
1

Hi, I’m a starter of this roboflow. I’m trying to use this for detecting Text box. First of all, I tried your AI but it couldn’t detect text area in my images(game image) so now I’m annotating it by myself(rectangular box). But even if i increased dataset 34 → 72 samples, in the test ROI(region of interest) is not detected properly. actually It’s same.
This procedure is not correct for this model or was i doing something wrong? please help…

Describe your question/issue here! (delete this when you post)

  • Project Type: Object detection
  • Operating System & Browser: Windows 11 / Chrome
  • Project Universe Link or Workspace/Project ID: tooltips
2

Hi @지훈_배

I’d love to investigate into this for you. I’m curious to see what sort of model metrics you are currently getting. For complex CV tasks, we may need more than 72 samples to get the type of performance you’d like. Some other recommendations I’d have would be to make sure you apply relevant augmentations to your data (perhaps different contrast levels or rotations). To help me triage further, would you mind if I access your workspace?