Imagine that we predicted 2 bounding boxes but we only have 1 ground truth object.
\nLet's name them bbox_A and bbox_B. bbox_A has 0.90 IoU while bbox_B has 0 IoU.
\nLet's suppose that the bbox_A has the highest score. This means that the bbox will be counted as \"11\" true positives (it s hit for all of the 11 thresholds).
\nWhat about the second box? Is it gonna be counted as 11 false positives (since we have 11 thresholds) or just 1 false positive?.
Imagine that we predicted 2 bounding boxes but we only have 1 ground truth object.
\nLet's name them bbox_A and bbox_B. bbox_A has 0.90 IoU while bbox_B has 0 IoU.
\nLet's suppose that the bbox_A has the highest score. This means that the bbox will be counted as \"11\" true positives (it s hit for all of the 11 thresholds).
\nWhat about the second box? Is it gonna be counted as 11 false positives (since we have 11 thresholds) or just 1 false positive?.
from evaluation page
\n\n", "rawMarkdown": "from evaluation page\n\n> The metric sweeps over IoU thresholds in the range of 0.3 to 0.8 with a step size of 0.05, calculating an F2 score at each threshold. For example, at a threshold of 0.5, a predicted object is considered a \"hit\" if its IoU with a ground truth object is at least 0.5.\n\n> A true positive is the first (in confidence order, see details below) submission box in a sample with an IoU greater than the threshold against an unmatched solution box.\n\n> Once all submission boxes have been evaluated, any unmatched submission boxes are false positives; any unmatched solution boxes are false negatives.\n\n> The final F2 Score is calculated as the mean of the F2 scores at each IoU threshold.", "votes": null }, { "id": "1651200", "postDate": "01/15/2022 16:02:23", "content": "The metric sweeps over IoU thresholds in the range of 0.3 to 0.8 with a step size of 0.05, calculating an F2 score at each threshold. For example, at a threshold of 0.5, a predicted object is considered a \"hit\" if its IoU with a ground truth object is at least 0.5.
\nA true positive is the first (in confidence order, see details below) submission box in a sample with an IoU greater than the threshold against an unmatched solution box.
\nOnce all submission boxes have been evaluated, any unmatched submission boxes are false positives; any unmatched solution boxes are false negatives.
\nThe final F2 Score is calculated as the mean of the F2 scores at each IoU threshold.
\n
Even starfish correctly predicted may be FN and FP how ? let's consider you predicted correctly starfish at IOU 0.6, your f2 metric will count it as FP and FN in (0.65 0.7 0.75 0.8)
", "rawMarkdown": "Even starfish correctly predicted may be FN and FP how ? let's consider you predicted correctly starfish at IOU 0.6, your f2 metric will count it as FP and FN in (0.65 0.7 0.75 0.8)", "votes": null } ], "comments": [ { "id": 1650754, "author_name": "imeintanis", "author_url": "", "post_date": "01/15/2022 09:49:44", "content": "from evaluation page
\n\n", "votes": null, "replies": [] }, { "id": 1651200, "author_name": "lukaszborecki", "author_url": "", "post_date": "01/15/2022 16:02:23", "content": "The metric sweeps over IoU thresholds in the range of 0.3 to 0.8 with a step size of 0.05, calculating an F2 score at each threshold. For example, at a threshold of 0.5, a predicted object is considered a \"hit\" if its IoU with a ground truth object is at least 0.5.
\nA true positive is the first (in confidence order, see details below) submission box in a sample with an IoU greater than the threshold against an unmatched solution box.
\nOnce all submission boxes have been evaluated, any unmatched submission boxes are false positives; any unmatched solution boxes are false negatives.
\nThe final F2 Score is calculated as the mean of the F2 scores at each IoU threshold.
\n
Even starfish correctly predicted may be FN and FP how ? let's consider you predicted correctly starfish at IOU 0.6, your f2 metric will count it as FP and FN in (0.65 0.7 0.75 0.8)
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1620452": "Imagine that we predicted 2 bounding boxes but we only have 1 ground truth object.\nLet's name them bbox_A and bbox_B. bbox_A has 0.90 IoU while bbox_B has 0 IoU.\nLet's suppose that the bbox_A has the highest score. This means that the bbox will be counted as \"11\" true positives (it s hit for all of the 11 thresholds).\nWhat about the second box? Is it gonna be counted as 11 false positives (since we have 11 thresholds) or just 1 false positive?.", "1650754": "from evaluation page\n\n> The metric sweeps over IoU thresholds in the range of 0.3 to 0.8 with a step size of 0.05, calculating an F2 score at each threshold. For example, at a threshold of 0.5, a predicted object is considered a \"hit\" if its IoU with a ground truth object is at least 0.5.\n\n> A true positive is the first (in confidence order, see details below) submission box in a sample with an IoU greater than the threshold against an unmatched solution box.\n\n> Once all submission boxes have been evaluated, any unmatched submission boxes are false positives; any unmatched solution boxes are false negatives.\n\n> The final F2 Score is calculated as the mean of the F2 scores at each IoU threshold.", "1651200": "Even starfish correctly predicted may be FN and FP how ? let's consider you predicted correctly starfish at IOU 0.6, your f2 metric will count it as FP and FN in (0.65 0.7 0.75 0.8)" }, "source": "meta" }