Inferior results produced on FGFA as compared to reported on Paper

[deeplearn9000]

Hello there,

Since there are no weights or model available, I try to reproduce the results by adopting the identical provided Config,

Contradictory to the 1.5 gain in mAP, I receive a gain of 0.2 points. It would be great if you share a trained model or explain if I am missing some Bells and Whistles.

The achieved results are mentioned below:

Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.511
Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=1000 ] = 0.780
Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=1000 ] = 0.575
Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.079
Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.250
Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.574
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.631
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=300 ] = 0.631
Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=1000 ] = 0.631
Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=1000 ] = 0.158
Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=1000 ] = 0.445
Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=1000 ] = 0.684