Inverted behaviour of `max_iou`?

#1
by kevinconka - opened

If I decrease the max_iou, I get less true positives:

image.png

image.png

image.png

Code to reproduce

from pprint import pprint
import evaluate
from seametrics.payload.processor import PayloadProcessor

payload = PayloadProcessor(
        dataset_name="SENTRY_VIDEOS_DATASET_QA",
        gt_field="ground_truth_det_fused_id",
        models=["ahoy_IR_b2_engine_3_6_0_49_gd81d3b63_oversea"],
        sequence_list=["Sentry_2023_02_08_PROACT_CELADON_@6m_MOB_2023_02_08_12_51_49"],
        # tags=["GT_ID_FUSION"],
        tracking_mode=True
   ).payload

module = evaluate.load("SEA-AI/user-friendly-metrics")
res = module._compute(payload, max_iou=0.9, recognition_thresholds=[0.3, 0.5, 0.8])
pprint(res)
SEA.AI org

motmetrics uses iou inverted idky. its stupid but we can just do 1-max_iou before sending it to the motmetrics. Would you agree with this?

SEA.AI org

i did 1-max_iou before sending it

Sign up or log in to comment