How mAP calculated in TLT

Accelerated Computing Intelligent Video Analytics TAO Toolkit
1

I have trained a model with detectnet_v2 and retinanet in TLT-V3 , where i got 5.0 and 12.0 mAP respectively.

detectnet_v2
eval_config {
average_precision_mode: INTEGRATE

retinanet
eval_config {
average_precision_mode: SAMPLE

Later i have trained a yolo_v4 model and got 0.6 mAP for 80 epochs .

yolo_v4
eval_config {
average_precision_mode: SAMPLE

Is the mAP calculate out of 1 for yolo_V4 and out of 100 for retinanent and detectnet_v2?

2

For detectnet_v2, if the mAP shows 0.6, that means 0.6%.
For yolo_v4 or retinanet, ssd, dssd, faster-rcnn, if the mAP shows 0.6, that means 60%.