The AP value is low

[1106X]

Test: [4990/5000] eta: 0:00:01 class_error: 90.00 loss: 18.1650 (18.6082) loss_bbox_dn: 0.0000 (0.0000) loss_giou_dn: 0.0000 (0.0000) loss_ce_dn: 0.0000 (0.0000) loss_ce: 0.9389 (0.9540) loss_bbox: 0.5480 (0.6824) loss_gio u: 0.9914 (1.0005) loss_ce_0: 0.9541 (0.9501) loss_bbox_0: 0.5749 (0.6957) loss_giou_0: 1.0708 (1.0115) loss_bbox_dn_0: 0.0000 (0.0000) loss_giou_dn_0: 0.0000 (0.0000) loss_ce_dn_0: 0.0000 (0.0000) loss_ce_1: 0.9386 (0.9535) loss_bbox_1: 0.5457 (0.6834) loss_giou_1: 1.0185 (1.0027) loss_bbox_dn_1: 0.0000 (0.0000) loss_giou_dn_1: 0.0000 (0.0000) loss_ce_dn_1: 0.0000 (0.0000) loss_ce_2: 0.9471 (0.9546) loss_bbox_2: 0.5535 (0.6815) loss_giou_2: 1.0 117 (1.0001) loss_bbox_dn_2: 0.0000 (0.0000) loss_giou_dn_2: 0.0000 (0.0000) loss_ce_dn_2: 0.0000 (0.0000) loss_ce_3: 0.9385 (0.9502) loss_bbox_3: 0.5266 (0.6842) loss_giou_3: 0.9924 (0.9998) loss_bbox_dn_3: 0.0000 (0.0000)
loss_giou_dn_3: 0.0000 (0.0000) loss_ce_dn_3: 0.0000 (0.0000) loss_ce_4: 0.9348 (0.9552) loss_bbox_4: 0.5235 (0.6833) loss_giou_4: 0.9887 (1.0005) loss_bbox_dn_4: 0.0000 (0.0000) loss_giou_dn_4: 0.0000 (0.0000) loss_ce_dn_4: 0.0000 (0.0000) loss_ce_interm: 0.9262 (0.9633) loss_bbox_interm: 0.6536 (0.7635) loss_giou_interm: 1.0469 (1.0381) loss_bbox_dn_unscaled: 0.0000 (0.0000) loss_giou_dn_unscaled: 0.0000 (0.0000) loss_ce_dn_unscaled: 0.0000 (0.0 000) loss_xy_dn_unscaled: 0.0000 (0.0000) loss_hw_dn_unscaled: 0.0000 (0.0000) cardinality_error_dn_unscaled: 0.0000 (0.0000) loss_ce_unscaled: 0.9389 (0.9540) class_error_unscaled: 90.0000 (74.5717) loss_bbox_unscaled: 0.1096 (0.1365) loss_giou_unscaled: 0.4957 (0.5002) loss_xy_unscaled: 0.0407 (0.0496) loss_hw_unscaled: 0.0669 (0.0869) cardinality_error_unscaled: 892.0000 (892.7283) loss_ce_0_unscaled: 0.9541 (0.9501) loss_bbox_0_unscaled: 0.1150 (0.1391) loss_giou_0_unscaled: 0.5354 (0.5058) loss_xy_0_unscaled: 0.0390 (0.0503) loss_hw_0_unscaled: 0.0789 (0.0889) cardinality_error_0_unscaled: 892.0000 (892.7283) loss_bbox_dn_0_unscaled: 0.0000 (0.0000) loss_giou_dn0 unscaled: 0.0000 (0.0000) loss_ce_dn_0_unscaled: 0.0000 (0.0000) loss_xy_dn_0_unscaled: 0.0000 (0.0000) loss_hw_dn_0_unscaled: 0.0000 (0.0000) cardinality_error_dn_0_unscaled: 0.0000 (0.0000) loss_ce_1_unscaled: 0.9386 (0.9535) loss_bbox_1_unscaled: 0.1091 (0.1367) loss_giou_1_unscaled: 0.5092 (0.5014) loss_xy_1_unscaled: 0.0399 (0.0495) loss_hw_1_unscaled: 0.0691 (0.0872) cardinality_error_1_unscaled: 892.0000 (892.7283) loss_bbox_dn_1_unscaled: 0. 0000 (0.0000) loss_giou_dn_1_unscaled: 0.0000 (0.0000) loss_ce_dn_1_unscaled: 0.0000 (0.0000) loss_xy_dn_1_unscaled: 0.0000 (0.0000) loss_hw_dn_1_unscaled: 0.0000 (0.0000) cardinality_error_dn_1unscaled: 0.0000 (0.0000) loss ce_2_unscaled: 0.9471 (0.9546) loss_bbox_2_unscaled: 0.1107 (0.1363) loss_giou_2_unscaled: 0.5059 (0.5000) loss_xy_2_unscaled: 0.0422 (0.0493) loss_hw_2_unscaled: 0.0728 (0.0870) cardinality_error_2_unscaled: 892.0000 (892.7283 ) loss_bbox_dn_2_unscaled: 0.0000 (0.0000) loss_giou_dn_2_unscaled: 0.0000 (0.0000) loss_ce_dn_2_unscaled: 0.0000 (0.0000) loss_xy_dn_2_unscaled: 0.0000 (0.0000) loss_hw_dn_2_unscaled: 0.0000 (0.0000) cardinality_error_dn_2_un scaled: 0.0000 (0.0000) loss_ce_3_unscaled: 0.9385 (0.9502) loss_bbox_3_unscaled: 0.1053 (0.1368) loss_giou_3_unscaled: 0.4962 (0.4999) loss_xy_3_unscaled: 0.0407 (0.0496) loss_hw_3_unscaled: 0.0641 (0.0872) cardinalityerror 3_unscaled: 892.0000 (892.7283) loss_bbox_dn_3_unscaled: 0.0000 (0.0000) loss_giou_dn_3_unscaled: 0.0000 (0.0000) loss_ce_dn_3_unscaled: 0.0000 (0.0000) loss_xy_dn_3_unscaled: 0.0000 (0.0000) loss_hw_dn_3_unscaled: 0.0000 (0.00 00) cardinality_error_dn_3_unscaled: 0.0000 (0.0000) loss_ce_4_unscaled: 0.9348 (0.9552) loss_bbox_4_unscaled: 0.1047 (0.1367) loss_giou_4_unscaled: 0.4943 (0.5002) loss_xy_4_unscaled: 0.0407 (0.0496) loss_hw_4_unscaled: 0.064 8 (0.0870) cardinality_error_4_unscaled: 892.0000 (892.7283) loss_bbox_dn_4_unscaled: 0.0000 (0.0000) loss_giou_dn_4_unscaled: 0.0000 (0.0000) loss_ce_dn_4_unscaled: 0.0000 (0.0000) loss_xy_dn_4unscaled: 0.0000 (0.0000) loss hw_dn_4_unscaled: 0.0000 (0.0000) cardinality_error_dn_4_unscaled: 0.0000 (0.0000) loss_ce_interm_unscaled: 0.9262 (0.9633) loss_bbox_interm_unscaled: 0.1307 (0.1527) loss_giou_interm_unscaled: 0.5235 (0.5191) loss_xy_interm_unscaled: 0.0430 (0.0531) loss_hw_interm_unscaled: 0.0882 (0.0996) cardinality_error_interm_unscaled: 892.0000 (892.7283) time: 0.1191 data: 0.0018 max mem: 11092 Test: [4999/5000] eta: 0:00:00 class_error: 100.00 loss: 18.4705 (18.6084) loss_bbox_dn: 0.0000 (0.0000) loss_giou_dn: 0.0000 (0.0000) loss_ce_dn: 0.0000 (0.0000) loss_ce: 0.9389 (0.9542) loss_bbox: 0.5905 (0.6824) loss_gi ou: 1.0163 (1.0004) loss_ce_0: 0.9541 (0.9502) loss_bbox_0: 0.6714 (0.6957) loss_giou_0: 1.1055 (1.0114) loss_bbox_dn_0: 0.0000 (0.0000) loss_giou_dn_0: 0.0000 (0.0000) loss_ce_dn_0: 0.0000 (0.0000) loss_ce_1: 0.9422 (0.9536) loss_bbox_1: 0.6317 (0.6834) loss_giou_1: 1.0207 (1.0026) loss_bbox_dn_1: 0.0000 (0.0000) loss_giou_dn_1: 0.0000 (0.0000) loss_ce_dn_1: 0.0000 (0.0000) loss_ce_2: 0.9515 (0.9547) loss_bbox_2: 0.6606 (0.6816) loss_giou_2: 1. 0299 (1.0000) loss_bbox_dn_2: 0.0000 (0.0000) loss_giou_dn_2: 0.0000 (0.0000) loss_ce_dn_2: 0.0000 (0.0000) loss_ce_3: 0.9479 (0.9503) loss_bbox_3: 0.5948 (0.6842) loss_giou_3: 1.0124 (0.9998) loss_bbox_dn_3: 0.0000 (0.0000) loss_giou_dn_3: 0.0000 (0.0000) loss_ce_dn_3: 0.0000 (0.0000) loss_ce_4: 0.9387 (0.9553) loss_bbox_4: 0.5942 (0.6832) loss_giou_4: 1.0102 (1.0004) loss_bbox_dn_4: 0.0000 (0.0000) loss_giou_dn_4: 0.0000 (0.0000) loss_ce_dn_4: 0.0000 (0.0000) loss_ce_interm: 0.9747 (0.9634) loss_bbox_interm: 0.6450 (0.7636) loss_giou_interm: 1.0772 (1.0381) loss_bbox_dn_unscaled: 0.0000 (0.0000) loss_giou_dn_unscaled: 0.0000 (0.0000) loss_ce_dn_unscaled: 0.0000 (0. 0000) loss_xy_dn_unscaled: 0.0000 (0.0000) loss_hw_dn_unscaled: 0.0000 (0.0000) cardinality_error_dn_unscaled: 0.0000 (0.0000) loss_ce_unscaled: 0.9389 (0.9542) class_error_unscaled: 88.8889 (74.5565) loss_bbox_unscaled: 0.118 1 (0.1365) loss_giou_unscaled: 0.5082 (0.5002) loss_xy_unscaled: 0.0407 (0.0496) loss_hw_unscaled: 0.0730 (0.0869) cardinality_error_unscaled: 895.0000 (892.7330) loss_ce_0_unscaled: 0.9541 (0.9502) loss_bbox_0_unscaled: 0.134 3 (0.1391) loss_giou_0_unscaled: 0.5528 (0.5057) loss_xy_0_unscaled: 0.0390 (0.0503) loss_hw_0_unscaled: 0.0875 (0.0889) cardinality_error_0_unscaled: 895.0000 (892.7330) loss_bbox_dn_0_unscaled: 0.0000 (0.0000) loss_giou_dn_0 _unscaled: 0.0000 (0.0000) loss_ce_dn_0_unscaled: 0.0000 (0.0000) loss_xy_dn_0_unscaled: 0.0000 (0.0000) loss_hw_dn_0_unscaled: 0.0000 (0.0000) cardinality_error_dn_0_unscaled: 0.0000 (0.0000) loss_ce_1_unscaled: 0.9422 (0.9536 ) loss_bbox_1_unscaled: 0.1263 (0.1367) loss_giou_1_unscaled: 0.5104 (0.5013) loss_xy_1_unscaled: 0.0384 (0.0495) loss_hw_1_unscaled: 0.0789 (0.0872) cardinality_error_1_unscaled: 895.0000 (892.7330) loss_bbox_dn_1_unscaled: 0 .0000 (0.0000) loss_giou_dn_1_unscaled: 0.0000 (0.0000) loss_ce_dn_1_unscaled: 0.0000 (0.0000) loss_xy_dn_1_unscaled: 0.0000 (0.0000) loss_hw_dn_1_unscaled: 0.0000 (0.0000) cardinality_error_dn_1_unscaled: 0.0000 (0.0000) loss _ce_2_unscaled: 0.9515 (0.9547) loss_bbox_2_unscaled: 0.1321 (0.1363) loss_giou_2_unscaled: 0.5149 (0.5000) loss_xy_2_unscaled: 0.0406 (0.0493) loss_hw_2_unscaled: 0.0783 (0.0870) cardinality_error_2_unscaled: 895.0000 (892.733 0) loss_bbox_dn_2_unscaled: 0.0000 (0.0000) loss_giou_dn_2_unscaled: 0.0000 (0.0000) loss_ce_dn_2_unscaled: 0.0000 (0.0000) loss_xy_dn_2_unscaled: 0.0000 (0.0000) loss_hw_dn_2_unscaled: 0.0000 (0.0000) cardinality_error_dn_2_u nscaled: 0.0000 (0.0000) loss_ce_3_unscaled: 0.9479 (0.9503) loss_bbox_3_unscaled: 0.1190 (0.1368) loss_giou_3_unscaled: 0.5062 (0.4999) loss_xy_3_unscaled: 0.0402 (0.0496) loss_hw_3_unscaled: 0.0766 (0.0873) cardinality_error _3_unscaled: 895.0000 (892.7330) loss_bbox_dn_3_unscaled: 0.0000 (0.0000) loss_giou_dn_3_unscaled: 0.0000 (0.0000) loss_ce_dn_3_unscaled: 0.0000 (0.0000) loss_xy_dn_3_unscaled: 0.0000 (0.0000) loss_hw_dn_3_unscaled: 0.0000 (0.0 000) cardinality_error_dn_3_unscaled: 0.0000 (0.0000) loss_ce_4_unscaled: 0.9387 (0.9553) loss_bbox_4_unscaled: 0.1188 (0.1366) loss_giou_4_unscaled: 0.5051 (0.5002) loss_xy_4_unscaled: 0.0407 (0.0496) loss_hw_4_unscaled: 0.07 37 (0.0870) cardinality_error_4_unscaled: 895.0000 (892.7330) loss_bbox_dn_4_unscaled: 0.0000 (0.0000) loss_giou_dn_4_unscaled: 0.0000 (0.0000) loss_ce_dn_4_unscaled: 0.0000 (0.0000) loss_xy_dn_4_unscaled: 0.0000 (0.0000) loss _hw_dn_4_unscaled: 0.0000 (0.0000) cardinality_error_dn_4_unscaled: 0.0000 (0.0000) loss_ce_interm_unscaled: 0.9747 (0.9634) loss_bbox_interm_unscaled: 0.1290 (0.1527) loss_giou_interm_unscaled: 0.5386 (0.5190) loss_xy_interm_unscaled: 0.0398 (0.0531) loss_hw_interm_unscaled: 0.0934 (0.0996) cardinality_error_interm_unscaled: 895.0000 (892.7330) time: 0.1168 data: 0.0018 max mem: 11092 Test: Total time: 0:12:17 (0.1476 s / it) Averaged stats: class_error: 100.00 loss: 18.4705 (18.6084) loss_bbox_dn: 0.0000 (0.0000) loss_giou_dn: 0.0000 (0.0000) loss_ce_dn: 0.0000 (0.0000) loss_ce: 0.9389 (0.9542) loss_bbox: 0.5905 (0.6824) loss_giou: 1.0163 (1.0004 ) loss_ce_0: 0.9541 (0.9502) loss_bbox_0: 0.6714 (0.6957) loss_giou_0: 1.1055 (1.0114) loss_bbox_dn_0: 0.0000 (0.0000) loss_giou_dn_0: 0.0000 (0.0000) loss_ce_dn_0: 0.0000 (0.0000) loss_ce_1: 0.9422 (0.9536) loss_bbox_1: 0.6 317 (0.6834) loss_giou_1: 1.0207 (1.0026) loss_bbox_dn_1: 0.0000 (0.0000) loss_giou_dn_1: 0.0000 (0.0000) loss_ce_dn_1: 0.0000 (0.0000) loss_ce_2: 0.9515 (0.9547) loss_bbox_2: 0.6606 (0.6816) loss_giou_2: 1.0299 (1.0000) los s_bbox_dn_2: 0.0000 (0.0000) loss_giou_dn_2: 0.0000 (0.0000) loss_ce_dn_2: 0.0000 (0.0000) loss_ce_3: 0.9479 (0.9503) loss_bbox_3: 0.5948 (0.6842) loss_giou_3: 1.0124 (0.9998) loss_bbox_dn_3: 0.0000 (0.0000) loss_giou_dn_3: 0 .0000 (0.0000) loss_ce_dn_3: 0.0000 (0.0000) loss_ce_4: 0.9387 (0.9553) loss_bbox_4: 0.5942 (0.6832) loss_giou_4: 1.0102 (1.0004) loss_bbox_dn_4: 0.0000 (0.0000) loss_giou_dn_4: 0.0000 (0.0000) loss_ce_dn_4: 0.0000 (0.0000)
loss_ce_interm: 0.9747 (0.9634) loss_bbox_interm: 0.6450 (0.7636) loss_giou_interm: 1.0772 (1.0381) loss_bbox_dn_unscaled: 0.0000 (0.0000) loss_giou_dn_unscaled: 0.0000 (0.0000) loss_ce_dn_unscaled: 0.0000 (0.0000) loss_xydn unscaled: 0.0000 (0.0000) loss_hw_dn_unscaled: 0.0000 (0.0000) cardinality_error_dn_unscaled: 0.0000 (0.0000) loss_ce_unscaled: 0.9389 (0.9542) class_error_unscaled: 88.8889 (74.5565) loss_bbox_unscaled: 0.1181 (0.1365) loss_g iou_unscaled: 0.5082 (0.5002) loss_xy_unscaled: 0.0407 (0.0496) loss_hw_unscaled: 0.0730 (0.0869) cardinality_error_unscaled: 895.0000 (892.7330) loss_ce_0_unscaled: 0.9541 (0.9502) loss_bbox_0_unscaled: 0.1343 (0.1391) loss_g iou_0_unscaled: 0.5528 (0.5057) loss_xy_0_unscaled: 0.0390 (0.0503) loss_hw_0_unscaled: 0.0875 (0.0889) cardinality_error_0_unscaled: 895.0000 (892.7330) loss_bbox_dn_0_unscaled: 0.0000 (0.0000) loss_giou_dn_0_unscaled: 0.0000 (0.0000) loss_ce_dn_0_unscaled: 0.0000 (0.0000) loss_xy_dn_0_unscaled: 0.0000 (0.0000) loss_hw_dn_0_unscaled: 0.0000 (0.0000) cardinality_error_dn_0_unscaled: 0.0000 (0.0000) loss_ce_1_unscaled: 0.9422 (0.9536) loss_bbox_1_uns caled: 0.1263 (0.1367) loss_giou_1_unscaled: 0.5104 (0.5013) loss_xy_1_unscaled: 0.0384 (0.0495) loss_hw_1_unscaled: 0.0789 (0.0872) cardinality_error_1_unscaled: 895.0000 (892.7330) loss_bbox_dn_1_unscaled: 0.0000 (0.0000) lo ss_giou_dn_1_unscaled: 0.0000 (0.0000) loss_ce_dn_1_unscaled: 0.0000 (0.0000) loss_xy_dn_1_unscaled: 0.0000 (0.0000) loss_hw_dn_1_unscaled: 0.0000 (0.0000) cardinality_error_dn_1_unscaled: 0.0000 (0.0000) loss_ce_2_unscaled: 0. 9515 (0.9547) loss_bbox_2_unscaled: 0.1321 (0.1363) loss_giou_2_unscaled: 0.5149 (0.5000) loss_xy_2_unscaled: 0.0406 (0.0493) loss_hw_2_unscaled: 0.0783 (0.0870) cardinality_error_2_unscaled: 895.0000 (892.7330) loss_bbox_dn_2 _unscaled: 0.0000 (0.0000) loss_giou_dn_2_unscaled: 0.0000 (0.0000) loss_ce_dn_2_unscaled: 0.0000 (0.0000) loss_xy_dn_2_unscaled: 0.0000 (0.0000) loss_hw_dn_2_unscaled: 0.0000 (0.0000) cardinality_error_dn_2_unscaled: 0.0000 (0 .0000) loss_ce_3_unscaled: 0.9479 (0.9503) loss_bbox_3_unscaled: 0.1190 (0.1368) loss_giou_3_unscaled: 0.5062 (0.4999) loss_xy_3_unscaled: 0.0402 (0.0496) loss_hw_3_unscaled: 0.0766 (0.0873) cardinality_error_3_unscaled: 895.0 000 (892.7330) loss_bbox_dn_3_unscaled: 0.0000 (0.0000) loss_giou_dn_3_unscaled: 0.0000 (0.0000) loss_ce_dn_3_unscaled: 0.0000 (0.0000) loss_xy_dn_3_unscaled: 0.0000 (0.0000) loss_hw_dn_3unscaled: 0.0000 (0.0000) cardinality error_dn_3_unscaled: 0.0000 (0.0000) loss_ce_4_unscaled: 0.9387 (0.9553) loss_bbox_4_unscaled: 0.1188 (0.1366) loss_giou_4_unscaled: 0.5051 (0.5002) loss_xy_4_unscaled: 0.0407 (0.0496) loss_hw_4_unscaled: 0.0737 (0.0870) cardi nality_error_4_unscaled: 895.0000 (892.7330) loss_bbox_dn_4_unscaled: 0.0000 (0.0000) loss_giou_dn_4_unscaled: 0.0000 (0.0000) loss_ce_dn_4_unscaled: 0.0000 (0.0000) loss_xy_dn_4_unscaled: 0.0000 (0.0000) loss_hw_dn_4_unscaled: 0.0000 (0.0000) cardinality_error_dn_4_unscaled: 0.0000 (0.0000) loss_ce_interm_unscaled: 0.9747 (0.9634) loss_bbox_interm_unscaled: 0.1290 (0.1527) loss_giou_interm_unscaled: 0.5386 (0.5190) loss_xy_interm_unscaled: 0.0398 (0.0531) loss_hw_interm_unscaled: 0.0934 (0.0996) cardinality_error_interm_unscaled: 895.0000 (892.7330) Accumulating evaluation results... DONE (t=25.48s). IoU metric: bbox Average Precision (AP) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50 | area= all | maxDets=100 ] = 0.001 Average Precision (AP) @[ IoU=0.75 | area= all | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.000 Average Precision (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.001 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 1 ] = 0.010 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets= 10 ] = 0.015 Average Recall (AR) @[ IoU=0.50:0.95 | area= all | maxDets=100 ] = 0.016 Average Recall (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.000 Average Recall (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.001 Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.027

[inSight-mk1]

看你项目是中文，就直接用中文说了。 和你一样的，我是自定义数据集，按coco格式转的，train时val结果是正常的，test就全零了，不知道哪里错了

[1106X]

看你项目是中文，就直接用中文说了。 和你一样的，我是自定义数据集，按coco格式转的，train时val结果是正常的，test就全零了，不知道哪里错了

我使用的是coco2017数据集 并且使用这个数据集在作者实验室前期工作DAB-DETR的时候一切正常 在这个工作中才出现了问题 并且我没更改任何设置 只对数据集路径、bs和numwork做了改变 我在train时val的结果也是这样 莫名其妙的错误

[Ferry7z]

请问解决了吗，我也遇到了同样的问题

[maichm]

训练自定义数据时，用resnet backbone似乎是正常的。同样的数据换swim backbone也出现了这个情况

[letmejoin]

遇到了同样的问题，训练自己数据，用cascade-rcnn-x101 AP50可以到80多，但是DINO相同的backbone，AP50只有40多，一直找不到啥原因

[maichm]

训练自定义数据时，用resnet backbone似乎是正常的。同样的数据换swim backbone也出现了这个情况

用了pretrain模型之后效果就提升了很多

[EddieEduardo]

请问这个转换的json文件是xyxy格式还是xywh格式呢？

[FengLi-ust]

train的时候发现val数据集上结果很差吗？在coco上能否复现我们的结果呢？

[Ghy1209]

请问解决了吗，我也遇到了同样的问题

你好，解决了吗，我也是这样，我在用focalnet-dino微调时出现ap很低的情况