To address your observation about lower model confidence detecting better, this is a common trade-off in object detection. Lower confidence thresholds will indeed detect more objects, but they may also increase false positives. Higher confidence thresholds will be more selective, potentially missing some objects but with higher certainty in the detections it does make.
References:
- Train YOLOv10 on Custom Dataset with Ultralytics - YouTube
- AttributeError: Can't get attribute 'v10DetectLoss' on <module 'ultralytics.utils.loss' > · Issue #13931 · ultralytics/ultralytics (github.com)
- notebooks/notebooks/train-yolov10-object-detection-on-custom-dataset.ipynb at main · roboflow/notebooks (github.com)
- 🚀😊 Brain Tumor Object Detection Using YOLOv10 | Python | 😊🚀 (youtube.com)
- Train a YOLOv10 Model on a Custom Dataset: A Step-by-Step Guide (youtube.com)
- How to Train YOLOv10 models on Your Custom Dataset (youtube.com)
- YOLOv10: How to Train for Object Detection on a Custom Dataset (youtube.com)
- TACO YOLOv8 Train and Predict w/wandb (kaggle.com)
best_yolov10_garbage_classification.pt - weights after YOLOv10 finetuned on the https://universe.roboflow.com/material-identification/garbage-classification-3/dataset/2 dataset.