BBox Detection Results Analysis Dashboard

paligemma2_3b_mix_448_detect_locations Analysis

Overview Metrics

Additional Metrics

Visualizations

Top Performing Categories (by AP)

Category	Average Precision
bear	0.721
cat	0.645
dog	0.601
giraffe	0.525
train	0.518
fire hydrant	0.508
airplane	0.498
laptop	0.492
frisbee	0.423
bus	0.403

Bottom Performing Categories (by AP)

Category	Average Precision
knife	0.054
car	0.052
broccoli	0.047
apple	0.039
skis	0.039
traffic light	0.029
backpack	0.027
handbag	0.017
book	0.016
dining table	0.005

IoU Threshold Impact

The IoU (Intersection over Union) threshold determines how strict the evaluation is. Higher thresholds require more precise bounding box alignment with ground truth.

Metric	IoU Threshold	Value
AP	0.50:0.95 (averaged)	0.205
AP50	0.50	0.307
AP75	0.75	0.218

paligemma2_3b_mix_448_detect_locations_perfect_classifier Analysis

Overview Metrics

Additional Metrics

Visualizations

Top Performing Categories (by AP)

Category	Average Precision
bear	0.708
cat	0.697
dog	0.688
train	0.603
airplane	0.573
laptop	0.571
fire hydrant	0.556
giraffe	0.543
microwave	0.495
bus	0.493

Bottom Performing Categories (by AP)

Category	Average Precision
carrot	0.073
car	0.072
boat	0.072
backpack	0.067
handbag	0.058
skis	0.058
apple	0.057
broccoli	0.047
traffic light	0.040
book	0.031

IoU Threshold Impact

The IoU (Intersection over Union) threshold determines how strict the evaluation is. Higher thresholds require more precise bounding box alignment with ground truth.

Metric	IoU Threshold	Value
AP	0.50:0.95 (averaged)	0.265
AP50	0.50	0.415
AP75	0.75	0.274

Model Comparison

Overall Performance

Processing Speed

Performance by Object Size

Detailed Metrics Comparison

Model	AP	AP50	AP75	Avg Time/Image
paligemma2_3b_mix_448_detect_locations	0.205	0.307	0.218	4.49s
paligemma2_3b_mix_448_detect_locations_perfect_classifier	0.265	0.415	0.274	5.18s