Could you provide the fine-tuned weights?

[CatfishW]

I finetuned your base-class weight on MS COCO dataset using RTX4070ti after 100 epochs with the batch size of 1 while other hyperparamter settings remain the same in accordance with the scripts you provided, but it does not reproduce to the accuracy to your essay out of the blue. The evaluation results for 10 shots are as follows:

Averaged stats: class_error: 76.92 loss: 25.7722 (22.3666) loss_ce: 2.4055 (1.9647) loss_bbox: 0.9593 (0.8582) loss_giou: 1.2056 (1.0262) loss_ce_0: 2.3303 (1.9005) loss_bbox_0: 0.8101 (0.7742) loss_giou_0: 1.0498 (1.0038) loss_ce_1: 2.3550 (1.8903) loss_bbox_1: 0.8683 (0.8341) loss_giou_1: 1.0159 (0.9948) loss_ce_2: 2.4136 (1.9047) loss_bbox_2: 0.8322 (0.8132) loss_giou_2: 1.0873 (0.9953) loss_ce_3: 2.2909 (1.8730) loss_bbox_3: 0.8844 (0.8044) loss_giou_3: 1.0809 (1.0009) loss_ce_4: 2.1820 (1.9072) loss_bbox_4: 0.8562 (0.8144) loss_giou_4: 1.0902 (1.0071) loss_ce_unscaled: 1.2028 (0.9824) class_error_unscaled: 50.0000 (43.5332) loss_bbox_unscaled: 0.1919 (0.1716) loss_giou_unscaled: 0.6028 (0.5131) cardinality_error_unscaled: 299.5000 (299.4213) loss_ce_0_unscaled: 1.1651 (0.9503) loss_bbox_0_unscaled: 0.1620 (0.1548) loss_giou_0_unscaled: 0.5249 (0.5019) cardinality_error_0_unscaled: 298.4375 (298.3096) loss_ce_1_unscaled: 1.1775 (0.9451) loss_bbox_1_unscaled: 0.1737 (0.1668) loss_giou_1_unscaled: 0.5080 (0.4974) cardinality_error_1_unscaled: 299.3750 (299.0845) loss_ce_2_unscaled: 1.2068 (0.9523) loss_bbox_2_unscaled: 0.1664 (0.1626) loss_giou_2_unscaled: 0.5437 (0.4976) cardinality_error_2_unscaled: 298.7500 (298.6726) loss_ce_3_unscaled: 1.1455 (0.9365) loss_bbox_3_unscaled: 0.1769 (0.1609) loss_giou_3_unscaled: 0.5405 (0.5004) cardinality_error_3_unscaled: 299.3125 (299.0845) loss_ce_4_unscaled: 1.0910 (0.9536) loss_bbox_4_unscaled: 0.1712 (0.1629) loss_giou_4_unscaled: 0.5451 (0.5035) cardinality_error_4_unscaled: 299.4375 (299.2578)

• Novel Categories:
  Accumulating evaluation results...
  DONE (t=1.25s).
  IoU metric: bbox
  Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.075
  Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.163
  Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.061
  Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.015
  Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.061
  Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.127
  Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.141
  Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.213
  Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.222
  Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.043
  Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.160
  Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.360

Thus, I'm wondering if you could provide your fine-tuned weights, and my email address is [email protected].

Looking forward to your reply.

[nanfangAlan]

Are you finetuned with 8 x GPUs?