The model gives performance metrics results for training dataset. Why cant we see the accuracy measures for test dataset? I tried without roboflow app as well, but the predict function in yolov8 can generate the confusion matrix, recall, precision, mAP for train and val and not test dataset why ?
Could you explain more about what you mean by accuracy measures?
If you are looking for mAPs for the test dataset, you can find that by clicking “More Metrics” in the trained version and navigating to the “Test Set” tab. (a note: this is not available for models which were uploaded, read below for how to do in a Colab.)
If you want to do this within a Colab (using YOLOv8 object detection as an example), you can use the code cell under the “Inference with Custom Model” section.
!yolo task=detect mode=predict model=/kaggle/input/mymoddd/bestt.pt conf=0.25 data=/kaggle/working/test/images
This particular code did give me the predicted images but not the performance metrics like - confusion matrix, mAP, PR curve, etc which we get for train and val dataset
My mistake, I sent the wrong code sample. It should be under the heading “Validate Custom Model”. The mAP, confusion matrix and other statistics should be found in the content/runs folder.
This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.